Diane C. Cowper, M.A., Joseph D. Kubal, M.A., Bruce A. Ripley, B.S.



The VA maintains the largest health care system under a single management structure in the United States. In a recent NEJM article, Ashton and colleagues found substantial geographic variation in hospital use for eight cohorts of VA patients (Ashton et. al, 1999).1 Building on this finding as a foundation, this paper examines the geographic variation in utilization rates for a specific procedure performed in the VA: radical prostatectomy. Using ArcView geographic information system software, we examine the variation both within the VA in different areas of the country, and also compare VA and non-VA rates for this procedure.


Prostate cancer is the most common malignancy in men over the age of fifty in the United States. This year, 180,000 American men will discover they have the disease. Although it is a common disease among this particular population, there is considerable debate and controversy over the treatment of early stage, slow, and moderately growing prostate cancer. The treatment course varies from "watchful waiting" to the invasive procedure radical prostatectomy (RP), which is the surgical removal of the prostate gland. Radiation therapy and hormonal therapy are also used in the treatment of this disease. There is little research evidence, however, to support one treatment regimen over another. There has been ongoing discussion in the medical literature over the past several years about the use of RP and, because of the differing views and care treatment of healthcare providers,2 numerous articles have documented wide geographic variation in the use of the procedure3,4 and changes in the pattern of prostate cancer care over time.5-11 In addition to geographic variation, several studies have found RP rates vary by racial group and socioeconomic status.12-16 Litwin et al. (1998) concluded RP represents "a significant burden on the federal health care dollar and does not appear to be as definitively curative as expected."6

Regional or geographic variation in the utilization of this procedure may occur for several reasons. To begin with, some areas of the country may have different diagnostic intensity when it comes to administering screening for prostate disease. Early-stage prostate cancer is often asymptomatic. Diagnosis of the disease is increasingly made through testing; more specifically, by screening for prostate-specific antigen (PSA). Since there is a great deal of regional variation in the frequency of use of this controversial screening test, there is, consequently, variation in the rate individuals are diagnosed and variation in when and how often men undergo surgery.17 An examination of the use of RP among Medicare beneficiaries before and after the introduction of prostate specific antigen testing (Lu-Yao et al., 1997) found that the rates of RP have steadily increased since 1984. A sharp increase in RP rates followed the institution of PSA testing after which a decrease, particularly among older age groups was evident. 18 Regional variation in the use of RP may also be due to gaps in medical knowledge and/or the lack of consensus about "best medical practice" in the treatment of prostate cancer. Until recently, there has been an absence of controlled clinical trials comparing the risks and benefits of surgery, radiation therapy, hormonal therapy, and watchful waiting. Three clinical trials are underway at the present time, but it will be several years before the results of the trials are known. 19 Geographic variations in RP use may also be due to patient preferences and choices in the type of treatment they receive.17,21 While patients rely on their health care providers to explain and quantify risks and benefits to each treatment course, some of the trade-offs involved in a surgical decision to remove the prostate (e.g., risk of impotence and/or urinary incontinence) can only be assessed by patients.

Research Questions/Objectives

The United States Department of Veterans Affairs (VA) maintains the largest health care system under a single management structure in the United States. Under its purview, the VA maintains 163 hospitals, over 500 Community Based Outpatient Clinics, 134 nursing homes, 40 domiciliaries, 163 hospital-based outpatient clinics and 4 independent outpatient clinics, and 4 mobile clinics.

In a recent New England Journal of Medicine article, Ashton and colleagues found substantial geographic variation in hospital use for eight cohorts of VA patients.1 Building on this finding as a foundation, this pilot study examines the geographic variation in the utilization rate of a high variation procedure: radical prostatectomy. Using geographic information system (GIS) software, we examine the geographic variation both within the VA in different areas of the country and also compare and contrast the VA patterns with Medicare rates for this procedure.

Past studies of radical prostatectomy have examined the utilization in either the Medicare population, the VA population, or specific geographic areas. This is one of the first investigations to make a national comparison of a specific surgical procedure in both the Medicare and VA patient population using Geographic Information System tools, comparable geographic units and time frames. We chose to look at this particular procedure for several reasons. To begin with, it is a high variation procedure and provides an opportunity to assess whether the variation in utilization of this surgery is due to regional differences in physician practice patterns or whether difference may be due to organizational practices within hospital systems (i.e., VA v. private). Secondly, the gender-specific procedure eliminates the differences between the Medicare and VA populations by focusing on males. Using the same logic as the Ashton research team, we expected to find that VA would show less geographic variation than the Medicare population.1 This hypothesis is based on the similarity in sociodemographics of the VA patient population and the highly centralized system of the VA where there are centrally mandated directives on delivery of patient care. Additionally, the physicians in the VA are salaried and, therefore, have little incentive to perform surgeries for financial gain.

Recent Studies in VA

Wilt et al. (1999) found that the number of RPs more than doubled between 1986 and 1996, and the rate of RP at VA Medical Centers per male VA user increased by 40%.4 After controlling for age and year, the utilization of RP in West North Central, Mountain, West South Central, and Pacific Census division was 70%, 14%, 10% and 8% higher, respectively, whereas the utilization of RP in New England, East North Central, and Mid-Atlantic divisions was 38%, 31%, and 25% lower respectively, than the rest of the nation. Geographic variation in utilization decreased during the period between 1986 and 1996, but a twofold difference in RP utilization in 1996 remained between high- and low-utilization divisions. Utilization of RP at VAMCs increased over time and varied across geographic areas. Mortality within 30 days was less than 1% and decreased with time. The researchers concluded that differences in utilization may be caused by uncertainty regarding the effectiveness of early detection and treatment of prostate cancer. Dr. Timothy Wilt, a VA physician at the Minneapolis VAMC and the University of Minnesota heads the Prostate Cancer Intervention Versus Observational Trial (PIVOT) research team. PIVOT is a randomized clinical trial designed to determine whether RP or expectant management provides superior length and quality of life for men with clinically localized prostate cancer.20 Results from this study are critical to determining the preferred therapy for clinically localized prostate cancer. Unfortunately, results from this study will not be forthcoming for several years.

Recent Findings from Non-VA Studies

Harlan et al. (1995) examined the geographic, age, and racial variation in the treatment of local/regional carcinoma of the prostate.3 Their study's findings included that the age-adjusted proportion of men, age 50 and older, who received RP increased sharply between 1984 and 1991. They also found that the choice of treatment varied widely by geographic regions. In 1991, the proportion of men with prostate cancer that received RP was highest in Utah and lowest in Connecticut among men with localized and regional disease. The Surveillance, Epidemiology, and End Results (SEER) data from the US National Cancer Institute show a clear trend toward more aggressive treatment for prostate cancer, especially RP. Adding to the controversy in RP utilization, these researchers also found the proportion of black men who receive RP was substantially lower than that of white men and this disparity does not appear to be changing. Black-white differences in the use of radical protatectomy have also been documented in a study of California men. Morris et al. (1999) concluded that black men are receiving less aggressive treatment when diagnosed with prostate cancer.16 Klabunde et al. (1998) also found racial differences in treatment of prostate cancer in elderly men, but cautioned that further research was needed to fully understand the reasons for racial differences and promote a more rational use of health care resources.13

On the other hand, Xia et al. (1998) in their examination of the trends in RP for benign prostatic hyperplasia (BPH) among black and white men in the United States from 1980 to 1994 demonstrated that the black/white differences in prostatectomy for BPH that were observed in the 1980s have disappeared in recent years. Furthermore, they found that the rates or RP have declined dramatically in all age- and race-specific groups. They concluded that, "further work is needed to determine whether this convergence in discharge rates is due to equalization of access to medical care or to differences in utilization of alternative therapies."5 Further, Polednak (1998) points to the need to consider both stage at diagnosis and socioeconomic status when trying to explain variation in the use of RP.14 While finding that the prevalence of RP was less frequent among blacks than whites, race was not a statistically significant independent predictor when age and poverty rate were included in logistic regression models.

This cursory overview of recent studies in RP illustrates the current debate and controversy surrounding the utilization of RP in the treatment of prostate cancer. The following section outlines the methodology by which we compared and contrasted the use of RP in the Medicare and VA patient populations.


Geographic Unit of Analysis. Rates of RP were aggregated to Hospital Referral Region (HRR), as defined in the Dartmouth Atlas of Health Care 1999.17 In the Atlas, HRRs are defined "by documenting where patients were referred for major cardiovascular surgical procedures and for neurosurgery. Each hospital service area was examined to determine where most of its residents went for these services. The result was the aggregation of the 3,436 hospital service areas into 306 hospital referral regions. Each hospital referral region had at least one city where both major cardiovascular surgical procedures and neurosurgery were performed" (p.294). The shape files for HRRs to display geographically the results were obtained from Esri's GIS Solution for Health CD-ROM (1998). Using the Zip Code crosswalk file, veterans' zip codes were translated into HRR codes.

Populations Under Investigation. The non-VA population under examination is the 1995-96 Medicare enrollee population. The VA population is comprised of all unique users of the VA health care system in calendar year 1995-96. At the time, VA did not have an "enrolled" population (this has subsequently changed beginning in 1998), so unique users for two years was the only alternative to achieve close comparability in the denominator.

Sources of RP Data.

Non-VA (Medicare). The rates of radical prostatectomy were obtained from the Dartmouth Atlas of Health Care 1999. As part of its product, the Atlas contains a diskette with data from the 1995-96 MEDPAR files from the Health Care Financing Administration (HCFA), Department of Health and Human Services. RP was defined as the International Classification of Disease, ICD-9 CM procedure code 60.5. The RP rates are calculated as the number of men undergoing this procedure per 1,000 male Medicare enrollees for two years. To obtain the raw number of patients undergoing RP, we divided the Medicare population by 1,000 and multiplied by the RP rate. Univariate statistics were performed and the mean RP rate was derived. The rates were geographically compared to the national average and mapped.

Department of Veterans Affairs (VA). The number of radical prostatectomy was obtained from the Department of Veterans Affairs' administrative inpatient database, the Patient Treatment File (PTF) - Surgery File. Because VA files are organized by fiscal year, we accessed the 1994, 1995, and 1996 files to obtain the number of RPs for calendar years 1995 and 1996. All male veterans who were 65 years of age and older (i.e., Medicare age-eligible) having RP (ICD-9 CM code 60.5) for cancer of the prostate (ICD-9 CM code 185) at VA hospitals during these calendar years were extracted. Unique users were also obtained from VA administrative databases; we extracted data from both inpatient and outpatient files. RP rates were calculated as the number of veterans undergoing this procedure per 1,000 unique users for the two-year time frame. RP rates by HRR were geographically compared to the national VA average and mapped.


Maps were generated using these data and the ArcView GIS software developed by Esri. Descriptive analyses (univariate and bivariate) were performed on the VA and non-VA RP data and the results of these analyses are presented in the next section.


Medicare Population.

During the 1995-96 time frame, there were 105,033 radical prostatectomies performed on the male Medicare population, translating into a national average rate of 1.9 surgeries per 1,000 enrollees. The range by Hospital Referral Region was from a low of .5 per 1,000 to a high of 4.7 per 1,000 (SD=.678). The "top ten" HRR for RP surgery were: Baton Rouge, LA (4.7); Hattiesburg, MS (4.2); Bend, OR (4.1); Billings, MT (4.0); Boulder, CO (3.8); Wichita Falls, TX (3.7); Kettering, OH (3.6); Muskegon, MI (3.6); Traverse, MI (3.4); and St. Petersburg, FL (3.4).

Map 1 illustrates the geographic variation in the use of RP by Hospital Referral Region for the male Medicare population. Highlights of the results include:

* 57 regions have rates that are 30% or more higher than the national average
* 74 regions have rates that are more than 25% below the national average
* Rates are high in the Northwest, the Mountain and Great Plains states, Michigan, and parts of Florida and Mississippi
* Rates are low in the Northeast, much of the Midwest, and in parts of Florida and Texas
* Rates of RP for prostate cancer varied by a factor of more than nine, from 0.5 per 1,000 in the Binghamton, New York, Tuscaloosa, Alabama, and Harlingen, Texas Hospital Referral Regions to 4.7 in the Baton Rouge, Louisiana Hospital Referral Region.

VA Population.

The data obtained from the VA PTF Surgery File showed there were 1,500 RPs performed on the male Medicare age-eligible veteran population in the 1995-96 time frame. This number translates into a national average rate of .968 surgeries per 1,000 unique male users 65 years of age and older. The range by Hospital Referral Region was from a low of 0 per 1,000 to a high of 5.9 per 1,000 (SD=.968189). The "top ten" HRR for RP surgery were: Bangor, ME (5.87); Portland, ME (5.24); Lebanon, NH (4.20); Longview, TX (3.91); Davenport, IA (3.71); Johnstown, PA (3.57); Victoria, TX (3.33); Bend, OR (3.26); Columbia, MO (3.05); and Shreveport, LA (2.95).

Map 2 illustrates the geographic variation in the use of RP by Hospital Referral Region for the older veteran male population. Highlights of the results include:

* 70 regions have no VA RP surgeries performed during this time frame
* Where RPs are performed in the VA, the range is from .14 per1,000 to 5.87 per 1,000 (SD=.83).
* 98 HRRs have rates 30% or more above the national average
* 133 regions have rates that are more than 25% below the national average
* Rates are highest in three HRRs in the Northeast: Bangor, Maine; Portland, Maine, and Lebanon, New Hampshire.
* Areas with no RPs are found in much of the Midwest, the Northeastern states of New Jersey, New York and Pennsylvania, the Mountain and Pacific states.
* Where RP for prostate cancer was performed, the rate varied by a factor of more than forty, from 0.14 per 1,000 in the Roanoke, VA Hospital Referral Region to 5.87 per 1,000 in the Bangor, ME Hospital Referral Region.

Similarities and Differences

A simple bivariate analysis on the correlation between RP rates in the VA v. Medicare population showed that the rates are positively correlated, but the Pearson's R (.097) is not statistically significant at the .05 level (p=.091). Only one HRR is on both the VA and non-VA "top ten" list for RPs: Bend, OR. This finding is interesting because, if the variation in the use of this procedure is based solely on physician practice patterns in a geographic area, one would expect to find that the rates in VA and non-VA health care settings to be very similar. Thus, the correlation between the two rates would be very high and statistically significant. It appears, therefore, that the variation in the use of this procedure encompasses more than just the geographic area of the country.

Both the VA and non-VA health care systems showed considerable variation in the utilization of RP surgery. We were surprised to find that the variation is actually greater in the VA than in the Medicare population. Additionally, while the number of RPs in VA represent a very small percentage of the procedures done in the HRR (mean=1.5%; SD=1.93), there are some areas where the percent is as high as 16% (Map 3). Hospital Referral Regions with 5 percent or more of the total RP surgeries done in the VA are: Boston, MA (16.4%), Savannah, GA (14.8%), Lebanon, NH (10.5%), Portland, ME (8.9%), Bangor, ME (8.2%), Gainesville, FL (8.0%), Chicago, IL (7.7%), Pueblo, CO (6.0%), Shreveport, LA (5.3%), Jonesboro, AR (5.1%) and Columbia, MO (5.0%).

Conclusions(The Example of Radical Prostatectomy)

While this pilot study is only a first step in assessing the similarity and difference between the VA and non-VA health care physician practice patterns, the results show that it is necessary to look at both systems to gain a complete view of the geographic variation and overall utilization rate of RP in older males. These patterns are helpful in providing a background for hypothesis-generating research. There are, however, many more steps that need to be taken to in order to draw concrete conclusions. To begin with, these data only span a two-year period. It would be helpful to have data at several different time points to ensure that the patterns that are portrayed are not simply outliers for a given point in time. Longitudinal, time-series data for both populations should be obtained and analyzed. Second, additional data are needed at the patient-level. The patterns presented in this paper are on all male Medicare and male VA Medicare age-eligible patients. Yet, past research has shown that the rates of RP vary by age, racial group, and socio-economic status. Therefore, more descriptive information about patients undergoing RP is necessary. Also, it would be helpful to have more data on patient severity and the stage of the prostate cancer. Other data that would be of interest in exploring RP differences in the VA and non-VA are outcome measures, including functional status, quality of life, re-hospitalization, post-operative complications, and 30-day mortality.

Third, this examination only focused on one surgical procedure. It would be interesting to look at other gender-specific procedure to see if the patterns that are seen for one procedure are mirrored in other surgeries. Finally, it would be of benefit to obtain data on the types of hospitals that are located in a particular HRR. Overlaying tertiary hospitals onto the maps would provide a great deal of insight into whether the "supply" was driving the "demand" for specific procedures. We intend to continue our investigations using these newly acquired tools, refine our methods in assessing practice patterns, and obtain additional data on patients undergoing specific surgeries.

Authors' Biosketches

Diane C. Cowper received her M.A. in Sociology at the University of Virginia, Charlottesville, VA. Ms. Cowper is a Health Research Scientist at the Midwest Center for Health Services and Policy Research, a VA HSR&D Center of Excellence. She became Co-Director (with Denise M. Hynes, Ph.D.) of the VA Information Resource Center (VIReC) in 1998. Ms. Cowper has a Visiting Scholar appointment with the Institute for Health Services Research and Policy Studies at Northwestern University, Chicago, Illinois. A Demographer by background, Diane's major research interests center on the residential relocation patterns of the veteran population, veterans' access to and utilization of VA health care services, and veterans' preferences and choice in health care providers.

Joseph D. Kubal, M.A. is presently Center Manager at the VA Information Resource Center (VIReC) based within the Midwest Center for Health Services and Policy Research (MCHSPR), Hines VA Hospital, Hines, IL. Mr. Kubal received his M.A. degree in geography from Chicago State University. His primary interests include medical geography, health care marketing research, and demography.

Bruce A. Ripley, B.S. is the Director of the Planning System Support Group, VHA Office of Policy and Planning, in Gainesville, FL. Mr. Ripley has been, and continues to be, in the forefront of bringing GIS applications into the Department of Veterans Affairs (VA). His primary interests center on veterans geographic access to VA health care services, as well as trends in service utilization, sociodemographics of VA's enrollee population, and forecasting service needs for the VA.


1) Ashton CM, Petersen NJ, Souchek J, Menke TJ, Yu HJ, Pietz K, Eigenbrodt ML, Barbour G, Kizer KW, Wray NP. Geographic variations in utilization rates in Veterans Affairs hospitals and clinics. N Engl J Med 1999 340(1):32-9.

2) Fowler FJ Jr., Bin L, Collins MM, Roberts RG, Oesterling JE, Wasson JH, Barry MJ. Prostrate Cancer screening and beliefs about treatment efficacy: A national survey of primary care physicians and urologists. The American Journal of Medicine 1998;104:526-532.

3) Harlan L, Brawley O, Pommerenke F, Wali P, Kramer B. Geographic, age, and racial variation in the treatment of local/regional carcinoma of the prostate. Journal of Clinical Oncology 1995;13(1):93-100.

4) Wilt TJ, Cowper DC, Gammack JK, Going DR, Nugent S, Borowsky SJ. An evaluation of radical prostatectomy at Veterans Affairs Medical Centers: Time trends and geographic variation in utilization and outcomes. Medical Care 1999;37(10):1046-1056.

5) Xia Z, Jacobsen SJ, Bergstralh EJ, Chute CG, Katusic AK, Lieber MM. Secular changes in radical prostatectomy utilization rates in Olmsted County, Minnesota 1980 to 1995. Journal of Urology 1998;159:904-908.

6) Litwin MS, Pasta DJ, Stoddard ML, Henning JM, Carrol PR. Epidemiological trends and financial outcomes in radical prostatectomy among Medicare beneficiaries, 1991 to 1993. The Journal of Urology 1998;160:445-448.

7) Mettlin C. Changes in patterns of prostate cancer care in the United States: Results of American College of Surgeons Commission Cancer Studies, 1974-1993. The Prostate 1997;33:221-226.

8) Imperato PJ, Nenner RP, Will TO. Trends in radical prostatectomy in New York State. American Journal of Medical Quality 1996;11(4):205-213.

9) Gee WF, Holtgrewe HL, Albertsen PC, Litwin MS, Manyak MJ, O'Leary MP, Painter MR. Practice trends in the diagnosis and management of prostate cancer in the United States. The Journal of Urology 1995; 154:207-208.

10) Post PN, Kil PJM, Hendrikx AJM, Poortmans PMP, Crommelin MA, Coebergh JW. Trend and variation treatment of localized prostate cancer in the Southern part of the Netherlands, 1988-1996. European Urology 1999;36:175-180.

11) Merrill RM. Changes in the use of radical prostatectomy for treating prostate cancer in the USA. The Lancet 1996;348:963-964.

12) Xia Z, Roberts RO, Schottenfeld D, Lieber MM, Jacobsen SJ. Trends in prostatectomy for benign prostatic hyperplasia among black and white men in the United States: 1980 to 1994. Adult Urology 1999;53(6):1154-1159.

13) Klabunde CN, Potosky AL, Harlan LC, Kramer BS. Trends and Black/White differences in treatment for Nonmetastatic Prostate Cancer. Medical Care 1998;36(9):1337-1348.

14) Polednak AP. Prostate cancer treatment in black and white men: The need to consider both stage at diagnosis and socioeconomic status. Journal of the National Medical Association 1998;90:101-104.

15) Imperato PJ, Nenner RP, Will TO. Radical porstatectomy: Lower rates among African American men. Journal of the National medical Association 1996; 88(9):589-594.

16) Morris CR, Snipes KP, Schlar R. Wright WE. Sociodemographic factors associated with prostatectomy and utilization and concorance with physician data query for prostate cancer (United States). Cancer Causes & Control 1999 10(6):503-11.

17) Wennberg JE et al. The Quality of Medical Care in the United States: A report on the Medicare Program. The Dartmouth Atlas of Health Care 1999/ The Center for the Evaluative Clinic Sciences, Dartmouth Medical School. AHA Press, 1999.

18) Lu-Yao GL, Friedman M, Yao SL. Use of radical prostatectomy among Medicare beneficiaries before and after the introduction of prostate specific antigen testing. The Journal of Urology 1997;157:2219-2222.

19) http://www.nytimes.com/library/national/science/health/050900hth-doctors.html 5/10/00.

20) Wilt TJ, Brawer MK. The prostate cancer intervention versus observation trial (PIVOT). Oncology 1997;11(8):1133-1143.

1) Frydenber M. Radical prostatectomy: What do the patients really think of surgery? MJA 1988; 168:477-78.

1 Cowper, Kubal & Ripley -