OBJECTIVE: Hysterectomy is the most common major surgical procedure performed in the United States for nonobstetric reasons. Although most hysterectomies include removal of the cervix, the rate of supracervical procedures has increased in recent years. To provide evidence about the outcomes of both types of hysterectomy, we conducted a randomized clinical trial of total (TAH) or supracervical (SCH) hysterectomy (the “TOSH” trial). We report here an analysis of 24-month resource use by patients in this trial.
METHODS: A randomized controlled trial was performed at 3 clinical centers to compare resources used by 120 patients who received a total or supracervical abdominal hysterectomy. Service use during a 24-month follow-up period was identified from medical and billing records and patient reports. Each service used was assigned a relative value, which was then converted into 2002 U.S. dollars.
RESULTS: Overall resource use was similar in the 2 study groups in the first 12 months after randomization (TAH $5,870; SCH $6,018; 95% confidence interval for difference −$960, $1,255; P < .79) and for the full 24 months (TAH $6,448; SCH $7,479; 95% confidence interval for difference −$533, $2,616; P < .20). In exploratory multivariable analyses, resource use was significantly associated with baseline body mass index greater than or equal to 35 kg/m2 ($8,440 versus $6,398, P = .02) and heavy bleeding ($7,550 versus $5,368, P = .02).
CONCLUSION: We conclude that the use of medical care resources over a 24-month period is comparable for total and supracervical hysterectomy. The association of a woman’s weight and bleeding pattern with subsequent resource use requires further investigation.
LEVEL OF EVIDENCE: I
The postsurgery use of medical care resources is comparable for patients who receive a total hysterectomy and those who receive a supracervical hysterectomy.
From the Departments of *Medicine, †Obstetrics, Gynecology and Reproductive Sciences, ‡Epidemiology and Biostatistics, and §Institute for Health Policy Studies, School of Medicine, University of California, San Francisco, California; ¶University of Alabama, Birmingham, Alabama; ∥University of Tennessee, Memphis, Tennessee; and *Wayne State University, Detroit, Michigan.
This project was supported by grant no. UO1 HS09478 from the Agency for Healthcare Research and Quality.
Reprints are not available. Address correspondence to: Jonathan Showstack, PhD, MPH, Institute for Health Policy Studies, University of California, San Francisco, 3333 California Street, Suite 265, San Francisco, CA 94118–1944; e-mail: email@example.com.
Received November 9, 2003. Received in revised form January 29, 2004. Accepted February 5, 2004.
Hysterectomy is the most common major surgical procedure performed in the United States for nonobstetric reasons. Approximately 600,000 hysterectomies are performed in this country each year, with the majority using the abdominal route.1,2 Before cervical cytology screening was introduced, gynecologists routinely performed a total abdominal hysterectomy (TAH), ie, the removal of both the corpus uteri and cervix uteri.3 Supracervical hysterectomy (SCH) was reserved for rare clinical circumstances in which cervical removal presented undue risks to the patient. Concerns over possible difficulties in sexual functioning, pelvic support, and urinary problems after removal of the cervix have fueled a rise in patient and clinician interest in SCH during the past decade.4–7
Between 1988 and 1998, the rate of TAH in Denmark fell by 38%, and the rate of SCH increased by 458%,8 and data from New York showed a similar pattern,9 yet SCH still accounts for only approximately 2% of the hysterectomies performed in the United States.1 A recent randomized study reported that, compared with TAH, SCH patients experienced more rapid recovery and fewer short-term complications, although SCH was also associated with infrequent cyclical bleeding and cervical prolapse.10
To provide evidence about the outcomes of both types of hysterectomy, we conducted a randomized clinical trial of total or supracervical hysterectomy (the “TOSH” trial). We report here the results of an analysis that addresses the economic question, “What was the total 24-month resource use by patients assigned randomly to either total or supracervical hysterectomy?”
The null hypothesis was that there would be no difference between randomized groups in the amounts or types of resources used during the 24 months after the date of randomization. The perspective of this analysis was relative resource use (not “costs” or “charges” to insurers, providers, or individual patients).
The basic structure and clinical outcomes of the TOSH trial have been reported, including a diagram of the flow of participants through the trial.11 At 2 years postrandomization, there were no significant differences in clinical outcomes between the 2 study groups.11 To summarize the study methods that are most relevant to the analyses reported in this paper, women aged 30–50 years were recruited from 2 principal groups: 1) premenopausal women aged more than 30 years who had symptomatic uterine leiomyomata (bleeding, pressure, or pain) and had decided with their gynecologist to undergo an abdominal hysterectomy, and 2) premenopausal women aged more than 30 years who presented to care for abnormal uterine bleeding, had a documented minimum 3-month trial of hormonal management, and had decided with their gynecologist to undergo abdominal hysterectomy. Potential participants had to be English-speaking, not pregnant, and of reproductive age without laboratory evidence of menopause, and have no plans to move and no condition likely to be fatal in 3 years. Subjects also could not have evidence of cancer, endocrinopathy, coagulopathy, or other conditions that had specific surgical or medical therapeutic indications, and they could not have tried medical therapies other than cyclic progestins.
One hundred twenty-five subjects were recruited between January 1998 and June 2000 at the University of Alabama, Birmingham, the University of Tennessee, Memphis, and Wayne State University, Detroit (a fourth site contributed clinical, but not economic, data to the study and was not included in the analyses reported here). The University of California, San Francisco, was the study coordinating center.
Subjects were included in the analyses reported here only if they attended at least 5 of the 8 possible follow-up interviews during the 24 months postrandomization. This criterion assured that the absence of any reported service use in a particular time-period was not due simply to the absence of a study follow-up visit; the criterion eliminated 2 TAH subjects and 3 SCH subjects. The resulting study cohort for the analyses reported here included 60 TAH subjects and 60 SCH subjects (Birmingham: TAH = 25, SCH = 26; Detroit: TAH = 10, SCH = 10; Memphis: TAH = 25, SCH = 24). For various reasons, 3 of the 120 eligible subjects (2 TAH and 1 SCH) did not have a hysterectomy.11 Sixty-four percent of subjects underwent hysterectomy within 1 week of randomization and 90% within 90 days of randomization.
All subjects provided informed consent to participate. The study’s methods and consent procedures were approved by the Institutional Review Boards at each of the study clinical sites and at the University of California, San Francisco. Interim monitoring to assess safety was carried out by an independent Data and Safety Monitoring Board.
Subjects were followed up for 24 months after randomization. Data were collected from subjects in structured interviews at the time of randomization and during in-person visits at 12 and 24 months, and by telephone at 3, 6, 9, 15, 18, and 21 months after randomization. Baseline data collection included sociodemographic characteristics and clinical information. Quality of life was measured by a battery of instruments derived from the Medical Outcomes Study,12 the Maine13 and Maryland14 Women’s Health Studies, and several newly created measures.15
To ensure completeness, resource use data were collected from multiple sources. Every 3 months participants were asked about the use of inpatient and outpatient services during the previous 3 months. Information collected included the reason for the service, the site of the service, and all diagnostic and therapeutic procedures that were completed during the visit or hospitalization. In addition, resource use data were abstracted from both inpatient and outpatient medical records; these data included Diagnosis-Related Groups (DRGs), International Classification of Diseases (9th revision) diagnosis and procedure codes, and Current Procedural Terminology (CPT) codes. Physician services included all outpatient office visits and procedures and all inpatient procedures. Not included were pharmacy data and inpatient consultations that did not involve a procedure. Discrepancies between the subjects’ self-reported data and data collected from medical records were noted and medical records searched again for specific instances of nonmatches, with medical record data used in the case of otherwise unresolved discrepancies.
Studies on the economics of hysterectomies provide an example of the variety of methods used, results obtained, and the consequent difficulty in interpretation (Table 1). 16–22 Note that neither charges nor costs are standardized as reported in the studies indicated in Table 1 and that data in these studies were derived from single institutions. We chose to study “resource use” rather than “charges” or “costs” because it avoids some of the measurement issues associated with charges and costs and provides a metric that is independent of study site and time period.23
Diagnosis-Related Groups and relative value units associated with CPT codes provide widely used metrics for the relative amount of resources used for a hospitalization (DRGs) and for physician services (relative value units). Both DRGs and relative value units assign relative weights or values to services based on the estimated average amount of resources used for a particular type of hospitalization (DRGs) or for a specific service provided by a physician (relative value units). By applying a dollar multiplier to a relative weight, a dollar value can be assigned to each type of service. Medicare uses DRG weights to calculate payments for hospitalizations and relative value units to calculate payment for physician services. For example, in a study of the economics of hysterectomy, investigators used Medicare physician reimbursement rules to estimate professional fees.20
A resource use history was constructed for each subject. Based on information collected from the medical record, each hospitalization was assigned a DRG, and each of more than 2,000 individual physician services was assigned a CPT code and its associated relative value unit. Because several data sources were used, records were reviewed individually and duplicate reports of services were eliminated. Each DRG and relative value unit was converted into 2002 U.S. dollars through the application of the amounts paid by Medicare for that hospitalization or service in 2002.24,25
For example, a hospitalization for a hysterectomy without coexisting conditions (DRG 359) has a relative weight of 0.8345, whereas a hysterectomy with 1 or more coexisting conditions (DRG 358) has a relative weight of 1.2295. The conversion factor (multiplier) for DRG weights used by Medicare in 2002 was $4,156.74 for hospitalizations in large urban areas (with adjustments for other factors, such as labor costs).24 Therefore, in 2002 Medicare would have paid approximately $3,469 (0.8345 × $4,156.74) for a hospitalization in a large urban area for a hysterectomy without coexisting conditions and $5,111 for a hysterectomy with coexisting conditions (1.2295 × $4,156.74).
Similarly, for physician services, an intermediate office visit for an established patient (CPT 99212) has a relative value unit of 1.0, whereas a total hysterectomy (CPT 58150) has an relative value unit of 24.67. The conversion factor used for CPT unit values used by Medicare in 2002 was $36.20.25 Thus, in 2002 Medicare would have paid $36.20 (1.0 × $36.20) for an intermediate office visit and $893 (24.67 × $36.20) for the physician’s fee for a total hysterectomy. (Note that the surgical fee also includes a presurgical visit and all surgical follow-up visits for 90 days after surgery.)
It is important to recognize that the choice of a particular conversion factor will not change the relative relationship among services (because a conversion factor is a constant that is applied to all services). The use of a dollar amount is, in fact, a convenience that allows the reporting of results in an understandable metric (dollars). We chose to use the Medicare conversion factors cited above because Medicare is a major national payer and accounts for a substantial portion of national hospital and physician reimbursement. The totals and averages produced using Medicare payment calculations, however, may appear to be low compared with the rates charged by hospitals and physicians. For example, if we had used the average charges at our study sites as the multiplier, the absolute amounts reported would be approximately 40% higher, but the relationships among the services and the interpretation of the relative resource use would not have changed.
The primary dependent variables were inpatient and outpatient resources used during the 24-month period after randomization. As an additional measure of resource use, we compared the average length of hospital stay for each group for the study hysterectomy. The analyses used the intention-to-treat principle without regard to treatment actually received. Because subjects were randomized to treatment, the main comparisons between groups were unadjusted. Two-tailed P values < .05 were considered statistically significant, with P values ≥ .05 and < .10 described as a “trend.”
To examine the relationship between subjects’ baseline demographic and clinical characteristics and resource use, we also performed exploratory, hypothesis-generating analyses using multiple linear regression. To minimize inflation of the type-I error rate, independent variables were added in prespecified blocks based on prior hypotheses.26
Included as independent variables in regressions models were study assignment; sociodemographic characteristics (age, weight [body mass index, BMI], being African American, being a high school graduate, and presence of health insurance); health-related quality-of-life measures (the Short Form Health Survey-36 [SF-36] physical component summary and mental component summary); symptoms at the time of randomization (duration of menstrual cycle [days], menstrual interval [days], an irregular menstrual cycle pattern, heavy or very heavy bleeding during menstrual cycle); clinical diagnoses at the time of the hysterectomy (the presence of uterine fibroids, pelvic pain or pressure); surgical pathology diagnoses (leiomyoma, adenomyosis, and endometritis [acute or chronic]); and having had antibiotic prophylaxis. Categorization of BMI as less than 35 kg/m2 or greater than or equal to 35 was based on exploratory data analysis.
The estimated number of subjects needed for the TOSH study was based on quality-of-life outcomes, not resource use outcomes. To indicate precision of estimated differences and assist in the interpretation of between-group comparisons that did not reach statistical significance, 95% confidence intervals (CI) are presented. Because the resource use outcome was skewed, with a few cases with very high resource use potentially affecting the power and validity of statistical tests, we performed 3 sensitivity analyses in addition to the primary analyses of all observed resource use values. First, we omitted cases that were more than 3 standard deviations (SDs) greater than the mean; second, we analyzed Winsorized27 outcomes, in which outliers omitted in the first sensitivity analysis were recoded to 3 SDs greater than the mean; and third, we analyzed the log transformation of resource use. Because all results were similar, we present only the results of the primary analysis using all of the untransformed resource use values.
Finally, selected postrandomization variables, including whether the subject had a hysterectomy during the study period and whether the subject was rehospitalized after the study hysterectomy, were added to the exploratory models to assess possible mediation of associations between baseline variables and resource use. Attenuation of coefficients for the baseline variables is interpretable as evidence for mediation.
There were no significant differences between the treatment groups in demographic characteristics, bleeding pattern, or preoperative diagnoses (Table 2). Average age was 41.8 years for TAH and 42.2 years for SCH (P = .62). Eighty-five percent of TAH subjects and 77% of SCH subjects were African American. Subjects averaged 83.9 kg in the TAH group and 82.8 kg in the SCH group (P = .79), and 28% of both TAH and SCH subjects (P = 1.00) had a BMI of 35 kg/m2 or greater.
The primary analyses showed that resource use was similar in the 2 study groups in the first 12 months after randomization (TAH $5,870; SCH $6,018; 95% CI for difference −$960, $1,255; P < .79) and for the full 24 months (TAH $6,448; SCH $7,479; 95% CI for difference −$553, $2,616; P < .20) (Table 3). Between-group differences in resource use were similar across the 3 study sites, as well as across subgroups defined by baseline BMI. Mean length of stay for the study hysterectomy was 3.6 days for TAH and 3.3 days for SCH (95% CI for difference 0.103, −0.747; P = .14). There were no readmissions for complications that resulted from the study hysterectomy. There was a trend for the SCH group to use more resources in the second follow-up year (P = .07) (Table 3), which appears to have been related to a higher rate of rehospitalization for the SCH group in the second year (15 rehospitalizations in the SCH group, 4 of which may have been hysterectomy-related; 4 rehospitalizations in the TAH group, none hysterectomy-related).
Among the baseline demographic and clinical variables, adjusted resource use was significantly associated only with BMI and heavy or very heavy bleeding during the menstrual cycle (Figure 1). Compared with subjects who had a BMI less than 35 kg/m2, subjects with a BMI of 35 kg/m2 or greater used 32% more resources (adjusted mean resource use $8,440 versus $6,398; P = .02). Subjects who had heavy or very heavy bleeding had an adjusted increase in 24-month resource use of 41% (adjusted mean resource use $7,550 versus $5,368; P = .02). We found an interaction between BMI of 35 kg/m2 or greater and having heavy bleeding, which in combination appeared to predict particularly high resource use. Both the main and interaction effects of high BMI and heavy bleeding were substantially mediated by rehospitalization after the study hysterectomy. Higher scores on the baseline SF-36 physical component summary significantly predicted lower subsequent resource use; a 1-point increase in the baseline physical component summary was associated with a decrease of $107 in 24-month resource use (P = .01). In contrast, we did not find an association of the mental health component summary of the SF-36 with resource use (P = .83). When comparisons of resource use were adjusted for posthysterectomy rehospitalization, the point estimate for the between-group difference in resource use in months 13–24 was eliminated almost entirely.
The most important economic finding of TOSH is that the total number of resources used by subjects who were randomized to either total or supracervical hysterectomy is comparable, especially in the first year. These results add substantially to previous literature in several important ways: 1) TOSH is the first randomized comparison between total abdominal and supracervical hysterectomy conducted in the United States; 2) the economic measure was resource use, rather than charges or costs, which allows a standardized comparison of the economic burden of treatment across sites and time; and 3) all inpatient and outpatient services for 24 months after the date of randomization were included.
There have been several studies of the costs of hysterectomy, almost all of which are observational and limited to the surgical hospitalization, and none have compared total abdominal hysterectomy with supracervical hysterectomy (Table 1). In general, the studies summarized in Table 1 show little difference in resource use between total abdominal hysterectomy and vaginal hysterectomy, although laparoscopically assisted vaginal hysterectomy appears to have been associated with somewhat higher costs.
The prevalence of obesity in the U.S. population has increased dramatically in recent years,28 and higher body weight has been shown to be associated with higher rates of morbidity and mortality.28,29 Other than an increase in wound infections, obesity has not been found to be strongly associated with perioperative outcomes of gynecologic or other types of surgery.30 Because a relatively high proportion of subjects in our study were obese, we were able to assess the association of obesity with long-term resource use after hysterectomy. The exploratory multivariate analyses suggested that a woman’s weight was strongly and independently associated with resource use. Women who had a BMI of 35 kg/m2 or greater, particularly those who had heavy bleeding at baseline, used significantly more resources over the 24 months than did women with lower BMIs.
Although there are several plausible clinical explanations for the association between bleeding and later resource use, including the impact of chronic bleeding and anemia on a woman’s overall health status, our study does not provide data to help distinguish among these possibilities. Other studies have shown that persons with a history of high resource use tend to continue to use more resources in future periods,31 perhaps because they may have greater needs and their prior experience may make them better able to enter and navigate the health care system. Although we do not have historical resource use data for the women in our study, it seems possible that those with a higher level of morbidity, such as heavy bleeding, at the time of entry to our study would have used more services in the past and therefore were likely to continue to use more services after their entry into our study.
The most important possible limitation to our study was that neither subjects nor clinicians were blind to a subject’s treatment assignment. It seems unlikely, however, that an awareness that a subject was assigned to either total or supracervical hysterectomy would have affected subsequent use of services substantially. Second, we did not have the statistical power to rule out differences in total resource as large as $2,616 over 2 years (Table 3); however, there were no rehospitalizations resulting from specific complications of the study hysterectomy, and any potential difference appears to be associated with hospitalizations in the supracervical group during the second year, most of which were unlikely to be related to the study hysterectomy. Third, although we measured many aspects of resource use comprehensively, we were not able to measure the number of in-hospital medical consultations or the use of pharmaceuticals. Because length of stay did not differ between the study groups, it seems unlikely that in-hospital medical consultations would have differed substantially. Pharmaceuticals account for approximately 10% of health care expenditures.32 There is little clinical reason to believe that there was a substantially different pattern of use of pharmaceuticals between the study groups. Finally, our findings for the associations of BMI and heavy bleeding history with resource use (P = .02 for both) were based on exploratory analysis and should be interpreted with caution.
We conclude that the use of medical care resources over a 24-month period appears comparable for total and supracervical hysterectomy. The association of a woman’s weight and bleeding pattern with subsequent resource use requires further investigation.
1. Farquhar CM, Steiner CA. Hysterectomy rates in the United States 1990–1997. Obstet Gynecol 2002;99:229–34.
2. Keshavarz H, Hillis SD, Kieke BA, Marchbanks PA. Hysterectomy surveillance—United States, 1994–1999. MMWR CDC Surveill Summ 2002;51(SS-5):1–7.
3. Munro MG. Supracervical hysterectomy: ...a time for reappraisal. Obstet Gynecol 1997;89:133–9.
4. Acken HS, Zarou GS. Hysterectomy for benign lesions. Am J Obstet Gynecol 1962;84:113–22.
5. Hanley HG. The late complications of total hysterectomy. Br J Urol 1969;41:682–4.
6. Vervest HA, Kiewiet de Jonge MK, Vervest TM, Barents JW, Haspels AA. Micturation symptoms and urinary incontinence after nonradical hysterectomy. Acta Obstet Gynecol Scand 1988;67:141–6.
7. Zussman L, Zussman S, Sunley R, Bjornson E. Sexual response after hysterectomy-oophorectomy: recent studies and reconsideration of psychogenesis. Am J Obstet Gynecol 1981;140:725–9.
8. Gimbel H, Settnes A, Tabor A. Hysterectomy on benign indication in Denmark 1988–1998: a register based trend analysis. Acta Obstet Gynecol Scand 2001;80:267–72.
9. Sills ES, Saini J, Applegate MS, McGee M, Gretz HF 3rd. Supracervical and total hysterectomy trends in New York State: 1990–1996. J Urban Health 1998;75:903–10.
10. Thakar R, Ayers S, Clarkson P, Stanton S, Manyonda I. Outcomes after total versus subtotal abdominal hysterectomy. N Engl J Med 2002;347:1318–25.
11. Learman L, Summitt RL, Varner E, McNeeley SG, Goodman-Gruen D, Richter HE, et al. A randomized comparison of total or supracervical hysterectomy: surgical complications and clinical outcomes. Obstet Gynecol 2003;102:453–62.
12. Stewart AL, Ware JE Jr, editors. Measuring functioning and well-being: the medical outcomes study approach. Durham (NC): Duke University Press; 1992.
13. Carlson KJ, Miller BA, Fowler FJ Jr. The Maine Women’s Health Study: I. Outcomes of hysterectomy. Obstet Gynecol 1994;83:556–65.
14. Rhodes JC, Kjerulff KH, Langenberg PW, Guzinski GM. Hysterectomy and sexual functioning. JAMA 1999;282:1934–41.
15. Varner RE, Ireland CC, Summit RL Jr, Richter HE, Learman LA, Vittinghoff E, et al, for the Ms Research Group. Medicine or surgery (Ms): a randomized trial comparing hysterectomy and medical treatment in premenopausal women with abnormal uterine bleeding. Available at: http://coordinatingcenter.ucsf.edu/ms
. Retrieved February 19, 2004.
16. Weber AM, Lee JC. Use of alternative techniques of hysterectomy in Ohio, 1988–1994. N Engl J Med 1996;335:483–9.
17. Johns DA, Carrera B, Jones J, DeLeon F, Vincent R, Safely C. The medical and economic impact of laparoscopically assisted vaginal hysterectomy in a large, metropolitan, not-for-profit hospital. Am J Obstet Gynecol 1995;172:1709–19.
18. Nezhat C, Bess O, Admon D, Nezhat AN, Nezhat F. Hospital cost comparison between abdominal, vaginal, and laparoscopy-assisted vaginal hysterectomies. Obstet Gynecol 1994;83:713–6.
19. Simon NV, Laveran RL, Cavanaugh S, Gerlach DH, Jackson JR. Laparoscopic supracervical vs. abdominal hysterectomy in a community hospital: a cost comparison. J Reprod Med 1999;44:339–45.
20. Dorsey JH, Holtz PM, Griffiths RI, McGrath MM, Steinberg EP. Costs and charges associated with three alternative techniques of hysterectomy. N Engl J Med 1996;335:476–82.
21. Summitt RL Jr, Stovall TG, Steege JF, Lipscomb GH. A multicenter randomized comparison of laparoscopically assisted vaginal hysterectomy and abdominal hysterectomy in abdominal hysterectomy candidates. Obstet Gynecol 1998;92:321–6.
22. Ellstrom M, Ferraz-Nunes J, Hahlin M, Olsson JH. A randomized trial with a cost consequence analysis after laparoscopic and abdominal hysterectomy. Obstet Gynecol 1998;91:30–4.
23. Katz P, Showstack J, Lake JR, Brown RS Jr, Dudley RA, Colwell ME, et al. Methods to estimate and analyze medical care resource use: an example from liver transplantation. Int J Technol Assess Health Care. 1999;15:366–79.
24. Federal Register. Hospital inpatient prospective payment systems and graduate medical education rates and costs (2002 FY). Balanced Budget Refinement Act provisions, etc. August 11, 2001. p. 39954, Table 1A. Available at: http://www.access.gpo.gov/su_docs/fedreg/a010801c.html
. Retrieved February 19, 2004.
26. Cohen J, Cohen P. Applied multiple regression/correlation analysis for the behavioral sciences. 2nd ed. Hillsdale (NJ): Lawrence Erlbaum Associates, Inc; 1983.
27. Ruppert D. Trimming and winsorization. In: Kotz S, Johnson NL, Read CB, editors. Encyclopedia of statistical sciences. Vol 9. New York (NY): John Wiley & Sons; 2003. p. 348–53.
28. Mokdad AH, Ford ES, Bowman BA, Dietz WH, Vinicor F, Bales VS, et al. Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001. JAMA 2003;289:76–9.
29. Manson JE, Willet WC, Stampfer MJ, Colditz GA, Hunter DJ, Hankinson SE, et al. Body weight and mortality among women. N Engl J Med 1995;333:677–85.
30. Flancbaum L, Choban PS. Surgical implications of obesity. Annu Rev Med 1998;49:215–34.
© 2004 The American College of Obstetricians and Gynecologists
31. Ash A, Porell F, Gruenberg L, Sawitz E, Beiser A. Adjusting Medicare capitation payments using prior hospitalization data. Health Care Financ Rev 1989;10:17–29.