Geographic and Socioeconomic Variation in the Treatment of Prostate Cancer
http://www.100md.com
《临床肿瘤学》
the Department of Urology, David Geffen School of Medicine
Departments of Health Services and Biostatistics, School of Public Health
Jonsson Comprehensive Cancer Center, University of California, Los Angeles, Los Angeles, CA
ABSTRACT
PURPOSE: Within a framework of quality assessment, primary treatment choice constitutes an indicator of quality of care. This study examines geographic and socioeconomic variation in the primary treatment of men with prostate cancer during the era of prostate-specific antigen testing.
METHODS: Using the National Cancer Institute's Surveillance, Epidemiology, and End Results public use data files, we identified men with localized/regional prostate cancer who underwent surgery, radiation therapy, or watchful waiting. We used the year 2000 US Census information to ascribe education and income levels to these men based on their county of residence and ethnicity.
RESULTS: Among the 96,769 men with localized/regional prostate cancer (during 1995 to 1999) who had sufficient information for analysis, we observed significant geographic variation nationwide in surgical, radiation, and watchful waiting treatment rates (P <.0015). Patterns noted 10 years ago, such as higher surgical rates in western regions, persisted. Ethnicity, income, and grade were all independently associated with primary treatment, or lack thereof. Blacks and low-income patients had the lowest rates of surgery and radiation. Grade was the best predictor of aggressive treatment.
CONCLUSION: Nonclinical factors, such as ethnicity and income, were associated with the use of watchful waiting rather than surgery or radiation in men with early-stage prostate cancer. These findings have implications for quality of care.
INTRODUCTION
Interest in measuring health care quality grew from a demand by the public and third-party payers for more efficient, cost-conscious health care that does not compromise quality.1 If a model of quality assessment is used that divides the delivery of health care services into structure, process, and outcome, primary treatment constitutes a process measure.2 Consequently, nonclinical factors (such as geographic region or socioeconomic status [SES]) associated with differences in primary treatment rates may also influence the quality of care.
Geographic location has been associated with variations in medical therapies for various medical conditions, prompting concern about the underlying rationale for clinical decisions.3-5 For example, analyses of radiation and surgery rates for prostate cancer, performed before the widespread use of prostate-specific antigen (PSA) testing and the advent of brachytherapy, showed that patients were two to four times more likely to receive radiation therapy or surgery, based solely on geographic region.6,7 Since these studies were published, the pervasiveness of PSA screening has resulted in a greater proportion of men presenting with organ-confined prostate cancer. At the same time, brachytherapy, an alternate form of radiation therapy for men with early-stage prostate cancer, has gained widespread popularity.8,9 Initial studies comparing efficacy among competing treatment modalities demonstrate comparable cancer control in the majority of patients, yet geographic variation persists in primary treatment choice.10-12
Several investigators have also examined the influence of SES on clinical factors in men with prostate cancer, but they have done so using surrogates for SES instead of a broadly conceived definition. For example, one study found lower SES, as determined by military rank, is associated with higher grade tumors, whereas another found men residing in census tracts with higher poverty rates received conservative treatment more often.13,14 Furthermore, low-income patients are more likely than high-income patients to undergo watchful waiting as primary therapy, suggesting that SES may contribute to the choice of primary treatment.15
Since the Institute of Medicine's 1996 Health Care Quality Initiative, interest in the science of quality of care assessment has increased. Although geographic variation itself does not constitute poor-quality care, this variation raises concern. Variation based on clinical parameters may indicate uncertainty in standard practice guidelines; variation based on sociodemographic characteristics indicates a differential utilization of available resources and, potentially, poor quality of care. Determining whether geographic, ethnic, and socioeconomic variations exist in the treatment of men with localized prostate cancer provides a foundation for improving quality of care for the large number of men facing this disease.
In this study, we set out to define target areas for quality improvement by assessing the influence of nonclinical factors on primary treatment rates. We used the Surveillance, Epidemiology, and End Results (SEER) public use files to identify a national sample of patients with early-stage prostate cancer, including those younger than 65 years of age who lack Medicare coverage. We hypothesized that in the PSA era, primary treatment rates would have become established locally but not nationally, and that socioeconomic factors would independently influence treatment rates.
METHODS
Data Sources
SEER is a population-based tumor registry maintained by the National Cancer Institute for 11 regions in the United States chosen to represent epidemiologically significant population subgroups.16 The 11 regions comprise five states (Connecticut, Hawaii, Iowa, New Mexico, and Utah) and six metropolitan areas (Detroit, San Francisco, Atlanta, Seattle, Los Angeles County, and San Jose-Monterey). These regions reflect the general population in terms of education and poverty level, although for privacy purposes these variables are not recorded for each individual.
Study Population
We identified all patients older than age 20 diagnosed between 1995 and 1999 with a stage of localized/regional prostate cancer. Beginning in 1995, SEER collapsed all prostate cancer patient categories defined as "confined entirely to the organ" or "extended beyond the limits of the organ of origin directly into surrounding tissue or lymph nodes" into the single category of localized/regional. For each patient, we identified the following variables: county of residence, age at diagnosis, race/ethnicity, date of diagnosis, tumor grade, site-specific surgery, radiation, and radiation sequence with surgery.
We defined primary treatment as surgery (coded as radical/total prostatectomy), radiation, or watchful waiting. Because our focus was on primary treatment, we classified patients who underwent radiation subsequent to radical prostatectomy only as surgery. For this analysis, radiation therapy included beam radiation, radioactive implants, or a combination of the two methods. To be classified as watchful waiting, we required both a surgical code of "no surgery" and a radiation code of "none." patients with treatment coded as "recommended, unknown if done" or "unknown" were excluded from analysis. Patients were also excluded if they had radiation before surgery (due to incomplete radiation data), incomplete ethnicity or grade information, or insufficient data to determine the treatment sequence.
We created two socioeconomic variables (income and education) using data from the year 2000 US Census. The US Census Bureau Web site provides a tool called the American FactFinder that allows the user to create specific queries from the Census 2000 summary file. We created custom tables identifying the median educational attainment and median annual household income in 1999 dollars for men older than 25 years. We further stratified men by specific ethnicity (black alone, American Indian/Alaskan Native, Asian alone, Hispanic, white alone not Hispanic, and other) residing in each of the corresponding SEER counties. The US Census counties matched the SEER code for county precisely. We ascribed the median education and income level from the census data to each individual analyzed from the SEER data based on his county of residence and ethnicity.
Analysis
Descriptive statistics are presented for demographic and clinical data. Multiple comparisons of primary treatment rates in each region referenced against the composite rate were compared using a Bonferroni correction with an overall error rate of 0.05. Similar comparisons were assessed at the intraregional level. We performed multinomial logistic regression analysis with watchful waiting as the referent group to investigate a link between the likelihood of primary treatment and demographic variables. In addition to age, we included geographic location and grade as covariates. Age remained a continuous variable in the model, whereas income was dichotomized at the lowest quartile of $40,216. We categorized ethnicity as white, black, Hispanic, and other; we collapsed education level into less than high school, high school graduate, and college educated. We categorized tumor grade as poorly, moderately, or well-differentiated, and specified Los Angeles as the referent SEER region. All analyses were conducted with SAS software, version 8 (SAS Institute, Cary, NC).
RESULTS
During 1995 to 1999, 102,970 patients had stage coded as localized/regional prostate cancer and coding that allowed clear delineation of treatment type. Of these, we could not ascertain ethnicity and grade in 2,118 and 4,234, respectively. Many patients had multiple missing variables, leaving 96,769 (94%) men for inclusion in the final multivariate analysis. We compared the 6,201 patients omitted from the final analysis to the original sample and did not find any statistical differences.
Table 1 lists patients' demographic and clinical characteristics, stratified by treatment modality. Patients undergoing surgery were significantly younger, were more likely to be white, and tended to have less than a high school education. Compared with those undergoing surgery, radiation patients tended to have a higher income and be older; those undergoing watchful waiting were also significantly older and had a greater likelihood of well-differentiated tumors. Patients in Los Angeles underwent surgery or watchful waiting at significantly higher rates than did patients in other regions, whereas those in Detroit had notably higher radiation rates.
The patterns reflecting these significant geographic variations are shown in Fig 1. Because of the large sample size, we were able to detect small differences between regional and composite proportions. Of the regional surgery proportions, only New Mexico and Seattle did not differ statistically from the composite proportion. Among the remaining nine regions, five (Atlanta, Iowa, Los Angeles, San Jose, and Utah) had proportions above the composite, whereas four (Connecticut, Hawaii, Detroit, and San Francisco) had proportions below the composite. All of the regional proportions for radiation differed significantly from the composite radiation proportion. Radiation proportions were higher than the composite in Connecticut, Detroit, San Francisco, Atlanta, and Seattle. The composite watchful waiting proportion differed significantly from the regional proportions in Atlanta, Connecticut, Detroit, New Mexico, San Jose, Seattle, and Utah. Connecticut, New Mexico, and Utah had higher watchful waiting proportions than either the composite proportion or the other regions. These differences persisted when patients were stratified by age to younger or older than 70.
Within-region analyses (data not shown) revealed more subtle variations. For all 11 regions, treatment rates were much more consistent intraregionally than inter-regionally. For example, within San Jose, only one county differed statistically from the regional average for watchful waiting and radiation, and none differed for surgery.
We constructed multivariate models to assess the association of education, income, and tumor grade with primary treatment rates, after adjusting for region (Table 2). The odds of blacks undergoing surgery rather than watchful waiting were half those of whites. The odds of lower income patients undergoing surgery were 20% less than for higher income patients. Tumor grade was the strongest predictor of surgical therapy: the less differentiated the tumor, the more likely the patient was to undergo surgery.
Similarly, the odds of blacks receiving radiation therapy instead of watchful waiting were 20% lower than for whites. Other ethnicity, which includes men of American Indian/Alaska Native, Chinese, Filipino, Japanese, or Hawaiian ancestry, was also associated with lower radiation rates when compared with watchful waiting rates in whites. The odds of lower income patients receiving radiation treatment were one fourth lower than for whites. In summary, these patients (lower income, blacks, and other ethnicities) had markedly lower rates of radiation therapy. Lastly, higher tumor grade predicted radiation over watchful waiting.
DISCUSSION
Widespread interest in measuring quality has emerged, as consumers and third-party payers increasingly try to identify high-quality providers and institutions.17,18 However, the determination of high-quality care is often subjective. In his model of quality assessment, Donabedian2 proposed measuring structure as available medical resources, process as technical aspects of care, and outcomes as patients' ultimate health status. This paradigm has become the basis for contemporary health care quality assessment. As investigators move toward developing process measures to improve quality of care, highly prevalent conditions are being targeted.19 We examined geographic, ethnic, and socioeconomic variation in the treatment of men with prostate cancer, a highly prevalent disease, and identified three critical findings.
First, in the 10 years since the original geographic variation studies in prostate cancer, there has been minimal convergence in treatment rates at the national level. Prior studies found the highest surgical rates in the west and the lowest rates in the northeast—a pattern we continue to see.6,20,21 Our findings likewise corroborate prior studies of the use of radiation therapy.7 Age has been implicated as a mitigating factor in previous investigations of variation in radiation or surgical treatment rates; however, we found variations after controlling for age.7,20 Although current research suggests clinical equipoise among prostate cancer treatments, providers appear to favor the modality associated with their specialty. For example, urologists tend to trust the results of radical retropubic prostatectomy over radiation therapy, perhaps because the long natural history of prostate cancer creates difficulties assessing the long-term cancer control of the more recent radiation therapies. In addition, our multivariate analysis suggests that even among radiation oncologists, geographic variation exists in practice patterns. The odds ratios for regional variation were two-fold higher in Hawaii and Atlanta, compared with Connecticut and Los Angeles. We found less surgical geographic variation, with fewer regions demonstrating significant variation, and those differing did so to a lesser degree.
At the intraregional level, treatment rates are much more uniform. This contrasts with other studies that demonstrated more uniform rates at the national level. These investigators hypothesized that larger areas were better able to distribute the variance uniformly. Thus, smaller regions demonstrated more variation.22,23 Perhaps in the PSA era, earlier diagnosis has produced a more clinically homogeneous patient population, leading to diminished variation. However, a Connecticut study of prostate cancer incidence during the PSA era found widespread variation in incidence rates across even small geographic areas. The authors postulated that the variation reflected differential screening practices by individual practitioners, and not differences in the underlying disease burden in the population.24 Similarly, we doubt that there are underlying differences in disease burden by region. Unlike the Connecticut study, our analysis found uniform prostate cancer treatment rates at the intraregional level, suggesting that local consensus may exist in the management of men with early-stage prostate cancer.
Although geographic variation originally was studied in terms of resource consumption, today the implications of geographic variation extend to quality of care.25 Geographic variation does not necessarily translate to poor quality of care. Because the competing modalities appear equally efficacious for most patients, treatment differences may be a result of available resources and not individual preference. However, an examination of Veterans Affairs Medical Centers in the contiguous United States found a similar pattern of geographic variation in surgical rates,26 suggesting that surgical variation persists even in the face of adequate resources. In fact, Ashton et al27 found that hospital resources, defined as number of beds, actually correlated with use. This pattern was consistent for eight different chronic diseases. In this era of evidence-based medicine, such variations speak to the need to determine which patients are best served by which treatment modalities. Trials such as the recently abandoned American College of Surgeons Oncology Group Z0070, a randomized, controlled trial comparing surgery with brachytherapy, are critical to the elucidation of this issue.
Second, ethnicity may be associated with treatment rates. Ethnicity and SES must be examined simultaneously because of the high prevalence of poverty among ethnic minorities. Intuitively, different cultural backgrounds could predispose patients to certain treatment preferences.28 Several studies have documented higher radiation rates in blacks, and focus group work in this population confirms a cultural bias toward "avoiding the knife."13,21,29
However, receiving radiation as treatment is different than receiving no treatment at all. Our data showed that blacks of comparable stage as that of whites or Hispanics were more likely to receive no treatment when compared with surgery or radiation therapy. This may suggest a lack of education regarding the natural history of prostate cancer and available treatment options. There is a tendency toward conservative treatment in blacks in the Medicare population, in whom more advanced age, comorbidity, and the indolent nature of prostate cancer may heavily influence the decision-making process.15 The role of education in the treatment decision is speculative; social support, access to care, transportation issues, or physician bias are other potential factors influencing the decision. The crucial question is whether the patients are involved in that decision-making process and actively choose watchful waiting or are steered on that course by their physician. Furthermore, watchful waiting suggests monitoring of a known condition. Shavers et al30 found that blacks and Hispanics were less likely than whites to receive active monitoring during a 60-month period, suggesting ethnic disparities in prostate cancer management.30 A more qualitative examination of patient and physician decision making would allow for exploration of these issues. Quality is not impaired but rather is preserved if patient participation and respect for cultural beliefs drive the ethnic disparity. However, it behooves researchers to undertake analyses providing evidence for these assertions.
Third, we found an association between low income and watchful waiting, independent of age, ethnicity, disease stage, and geographic region. Although ethnic minorities tend to occupy a lower socioeconomic position, blacks and those of low income are predisposed to watchful waiting. Analysis of the Connecticut Tumor Registry supports the concept that poverty inversely predicts radical prostatectomy rates.31 In fact, when stage and poverty level were controlled for, ethnicity was no longer a significant predictor of treatment. The Connecticut Tumor Registry included a poverty variable at the census tract level. Klabunde et al13 confirmed that men residing in higher poverty regions, as determined by census tract, had lower surgery rates. These studies provide external validity for our use of a separate area-based measure of SES, namely the median income at the county level. We recognize that all of these measures are proxies for individual SES; however, because public health surveillance systems have not routinely collected individual data, researchers must use such proxies until such data become available.
Tarman et al14 used military rank as a surrogate for SES and found that patients of lower SES had significantly higher Gleason grade at presentation; thus, low-income patients are potentially most at risk from their disease. We found that the strongest predictor of surgery or radiation therapy was a moderate- or high-grade tumor, yet the low-income patients in our study more often received no treatment, after adjusting for grade.
Although those of lower income may have lacked adequate access to health care, they must have had access to some portion of the healthcare system to be included in the tumor registry. From our data set, it is unclear if the treatment disparity is due to poor quality of care within the health plan or access to a lower quality health plan. Analyses of a variety of medical activities including breast cancer screening, eye examinations in diabetics, knee arthroplasty, and the use of beta-blockers have found that within health plans, patients of different ethnicities are treated equally. However, socioeconomic variation persisted across samples because lower income patients were more commonly enrolled in lower quality health plans.5,32
The finding that lower income predisposes patients to watchful waiting as a therapeutic option has implications for quality of care, regardless of whether it occurs at the provider or health plan level. If the regional availability of medical resources influences the decision, it is hard to fathom why this would preferentially affect low-income patients. Carlisle et al4 demonstrated that ethnicity and income are more important determinants of surgical treatment rates than the availability of medical services. This may be the more concerning quality problem to combat. Provider-propagated differences may be implicated by our finding that education level was not associated with treatment rates. Less-educated patients may have a worse understanding of the nuances of treatment and difficulty comprehending the available literature, thus they might make less-informed choices.33,34 However, education did not affect treatment rates in our patients. Conversely, low-income patients may have comorbid conditions that preclude providers from offering them aggressive therapy. Another potential reason for the disparity in treatment options beyond the provider's control relates to hidden economic costs. Low-income patients may opt for less aggressive treatment because they perceive that they cannot afford time off from work or they have transportation issues.
Our study has several limitations. First, we had to ascribe group characteristics to individuals, which may not accurately reflect the income or education of that individual. However, given that public health surveillance systems lack individualized data, proxies must be used. Leading researchers in geocoding have found that area-based geosocial measures can be applied validly to individuals residing in that area.35,36 Many prostate cancer researchers have used SEER-Medicare–linked databases, which provide socioeconomic data at the census tract level. Although there are advantages to this approach, it fails to identify treatment patterns in younger men diagnosed with prostate cancer. Our analysis is particularly robust because it included men of all ages. Those younger than Medicare age are increasingly being diagnosed as PSA screening becomes more ubiquitous. Second, we did not have a detailed assessment of patients' health status. Although we hypothesized that men younger than 70 years of age with local/regional disease are comparable, this may not be the case. Comorbid conditions often drive therapy considerations in prostate cancer patients. Third, no data on patient preference or provider characteristics are available through SEER, and this could be an important underlying reason for the variations we observed. The characteristics of physicians caring for the lower income patients may differ from those of other physicians in terms of board certification, fellowship training, or practice duration, and may influence quality of care. Fourth, SEER limits data collection to 4 months from initial diagnosis. The inefficiencies of health systems serving the disadvantaged could lead to treatment delays extending beyond 4 months. Therefore, our data for the men of lower SES may not accurately reflect treatment received. However, this delay itself may negatively influence quality; researchers from Canada have found that a delay of 3 months may adversely affect cancer control.37 Lastly, we infer that different treatment rates suggest inequities in care. These treatment differences may be just. The grouping of localized disease with regional disease potentially confounds this issue because low-income patients are more likely to present with more advanced disease. This would predispose them to radiation, but not watchful waiting. Clinical parameters or underlying health status, not SES or ethnicity, should be the reason for the inequality.
Geographic, ethnic, and socioeconomic variations in prostate cancer treatment persisted into the late 1990s. These findings have implications for quality of care in that they may suggest unconscious discrimination on the part of physicians. Systematic studies of outcomes in this population, controlling for the confounding variables of ethnicity and SES, will help answer these questions. With better evidence to identify which patient populations benefit from each treatment modality, we will be able to disseminate these findings nationwide to improve quality of care for men with prostate cancer.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported by a grant from the American Foundation for Urologic Disease/American Urological Association Research Scholar Program.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Christianson R: ACHA 1998 Quality Award winners: Quality initiatives that stand out. Provider 24:81-89, 1998
Donabedian A: Quality assurance: Structure, process and outcome. Nurs Stand 7:4-5, 1992
Wennberg J, Gittelsohn A: Variations in medical care among small areas. Sci Am 246:120-134, 1982
Carlisle DM, Valdez RB, Shapiro MF, et al: Geographic variation in rates of selected surgical procedures within Los Angeles County. Health Serv Res 30:27-42, 1995
Skinner J, Weinstein JN, Sporer SM, et al: Racial, ethnic, and geographic disparities in rates of knee arthroplasty among Medicare patients. N Engl J Med 349:1350-1359, 2003
Lu-Yao GL, McLerran D, Wasson J, et al: An assessment of radical prostatectomy: Time trends, geographic variation, and outcomes—The Prostate Patient Outcomes Research Team. JAMA 269:2633-2636, 1993
Lai S, Lai H, Lamm S, et al: Radiation therapy in non-surgically-treated nonmetastatic prostate cancer: Geographic and demographic variation. Urology 57:510-517, 2001
Smith RP, Malkowicz SB, Whittington R, et al: Identification of clinically significant prostate cancer by prostate-specific antigen screening. Arch Intern Med 164:1227-1230, 2004
Hugosson J, Aus G, Lilja H, et al: Results of a randomized, population-based study of biennial screening using serum prostate-specific antigen measurement to detect prostate carcinoma. Cancer 100:1397-1405, 2004
Sylvester JE, Blasko JC, Grimm PD, et al: Ten-year biochemical relapse-free survival after external beam radiation and brachytherapy for localized prostate cancer: The Seattle experience. Int J Radiat Oncol Biol Phys 57:944-952, 2003
Pound CR, Partin AW, Eisenberger MA, et al: Natural history of progression after PSA elevation following radical prostatectomy. JAMA 281:1591-1597, 1999
D'Amico AV, Whittington R, Malkowicz SB, et al: Biochemical outcome after radical prostatectomy or external beam radiation therapy for patients with clinically localized prostate carcinoma in the prostate specific antigen era. Cancer 95:281-286, 2002
Klabunde CN, Potosky AL, Harlan LC, et al: Trends and black/white differences in treatment for nonmetastatic prostate cancer. Med Care 36:1337-1348, 1998
Tarman GJ, Kane CJ, Moul JW, et al: Impact of socioeconomic status and race on clinical parameters of patients undergoing radical prostatectomy in an equal access health care system. Urology 56:1016-1020, 2000
Shavers VL, Brown ML, Potosky AL, et al: Race/ethnicity and the receipt of watchful waiting for the initial management of prostate cancer. J Gen Intern Med 19:146-155, 2004
Surveillance, Epidemiology, and End Results (SEER) Program Public-Use Data (1973-2001): National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, 2003, based on November 2002 submission. http://www.seer.cancer.gov
Brook RH, McGlynn EA, Shekelle PG: Defining and measuring quality of care: A perspective from US researchers. Int J Qual Health Care 12:281-295, 2000
Birkmeyer JD, Dimick JB: Potential benefits of the new Leapfrog standards: Effect of process and outcomes measures. Surgery 135:569-575, 2004
Spencer BA, Steinberg M, Malin J, et al: Quality-of-care indicators for early-stage prostate cancer. J Clin Oncol 21:1928-1936, 2003
Lai S, Lai H, Krongrad A, et al: Radical prostatectomy: Geographic and demographic variation. Urology 56:108-115, 2000
Harlan L, Brawley O, Pommerenke F, et al: Geographic, age, and racial variation in the treatment of local/regional carcinoma of the prostate. J Clin Oncol 13:93-100, 1995
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 307:1310-1314, 1982
McPherson K, Strong PM, Epstein A, et al: Regional variations in the use of common surgical procedures: Within and between England and Wales, Canada and the United States of America. Soc Sci Med [A] 15:273-288, 1981
Gregorio DI, Kulldorff M, Sheehan TJ, et al: Geographic distribution of prostate cancer incidence in the era of PSA testing, Connecticut, 1984 to 1998. Urology 63:78-82, 2004
Wennberg JE, Freeman JL, Culp WJ: Are hospital services rationed in New Haven or over-utilised in Boston? Lancet 1:1185-1189, 1987
Wilt TJ, Cowper DC, Gammack JK, et al: An evaluation of radical prostatectomy at Veterans Affairs Medical Centers: Time trends and geographic variation in utilization and outcomes. Med Care 37:1046-1056, 1999
Ashton CM, Petersen NJ, Souchek J, et al: Geographic variations in utilization rates in Veterans Affairs hospitals and clinics. N Engl J Med 340:32-39, 1999
Knight SJ, Siston AK, Chmiel JS, et al: Ethnic variation in localized prostate cancer: A pilot study of preferences, optimism, and quality of life among black and white veterans. Clin Prostate Cancer 3:31-37, 2004
Sohayda CJ, Kupelian PA, Altsman KA, et al: Race as an independent predictor of outcome after treatment for localized prostate cancer. J Urol 162:1331-1336, 1999
Shavers VL, Brown M, Klabunde CN, et al: Race/ethnicity and the intensity of medical monitoring under ‘watchful waiting’ for prostate cancer. Med Care 42:239-250, 2004
Polednak AP: Prostate cancer treatment in black and white men: The need to consider both stage at diagnosis and socioeconomic status. J Natl Med Assoc 90:101-104, 1998
Schneider EC, Leape LL, Weissman JS, et al: Racial differences in cardiac revascularization rates: Does "overuse" explain higher rates among white patients? Ann Intern Med 135:328-337, 2001
Kane CJ, Lubeck DP, Knight SJ, et al: Impact of patient educational level on treatment for patients with prostate cancer: Data from CaPSURE. Urology 62:1035-1039, 2003
Kim SP, Knight SJ, Tomori C, et al: Health literacy and shared decision making for prostate cancer patients with low socioeconomic status. Cancer Invest 19:684-691, 2001
Krieger N, Williams DR, Moss NE: Measuring social class in US public health research: Concepts, methodologies, and guidelines. Annu Rev Public Health 18:341-378, 1997
Krieger N: Overcoming the absence of socioeconomic data in medical records: Validation and application of a census-based methodology. Am J Public Health 82:703-710, 1992
Nam RK, Jewett MA, Krahn MD, et al: Delay in surgical therapy for clinically localized prostate cancer and biochemical recurrence after radical prostatectomy. Can J Urol 10:1891-1898, 2003(Tracey L. Krupski, Lorna )
Departments of Health Services and Biostatistics, School of Public Health
Jonsson Comprehensive Cancer Center, University of California, Los Angeles, Los Angeles, CA
ABSTRACT
PURPOSE: Within a framework of quality assessment, primary treatment choice constitutes an indicator of quality of care. This study examines geographic and socioeconomic variation in the primary treatment of men with prostate cancer during the era of prostate-specific antigen testing.
METHODS: Using the National Cancer Institute's Surveillance, Epidemiology, and End Results public use data files, we identified men with localized/regional prostate cancer who underwent surgery, radiation therapy, or watchful waiting. We used the year 2000 US Census information to ascribe education and income levels to these men based on their county of residence and ethnicity.
RESULTS: Among the 96,769 men with localized/regional prostate cancer (during 1995 to 1999) who had sufficient information for analysis, we observed significant geographic variation nationwide in surgical, radiation, and watchful waiting treatment rates (P <.0015). Patterns noted 10 years ago, such as higher surgical rates in western regions, persisted. Ethnicity, income, and grade were all independently associated with primary treatment, or lack thereof. Blacks and low-income patients had the lowest rates of surgery and radiation. Grade was the best predictor of aggressive treatment.
CONCLUSION: Nonclinical factors, such as ethnicity and income, were associated with the use of watchful waiting rather than surgery or radiation in men with early-stage prostate cancer. These findings have implications for quality of care.
INTRODUCTION
Interest in measuring health care quality grew from a demand by the public and third-party payers for more efficient, cost-conscious health care that does not compromise quality.1 If a model of quality assessment is used that divides the delivery of health care services into structure, process, and outcome, primary treatment constitutes a process measure.2 Consequently, nonclinical factors (such as geographic region or socioeconomic status [SES]) associated with differences in primary treatment rates may also influence the quality of care.
Geographic location has been associated with variations in medical therapies for various medical conditions, prompting concern about the underlying rationale for clinical decisions.3-5 For example, analyses of radiation and surgery rates for prostate cancer, performed before the widespread use of prostate-specific antigen (PSA) testing and the advent of brachytherapy, showed that patients were two to four times more likely to receive radiation therapy or surgery, based solely on geographic region.6,7 Since these studies were published, the pervasiveness of PSA screening has resulted in a greater proportion of men presenting with organ-confined prostate cancer. At the same time, brachytherapy, an alternate form of radiation therapy for men with early-stage prostate cancer, has gained widespread popularity.8,9 Initial studies comparing efficacy among competing treatment modalities demonstrate comparable cancer control in the majority of patients, yet geographic variation persists in primary treatment choice.10-12
Several investigators have also examined the influence of SES on clinical factors in men with prostate cancer, but they have done so using surrogates for SES instead of a broadly conceived definition. For example, one study found lower SES, as determined by military rank, is associated with higher grade tumors, whereas another found men residing in census tracts with higher poverty rates received conservative treatment more often.13,14 Furthermore, low-income patients are more likely than high-income patients to undergo watchful waiting as primary therapy, suggesting that SES may contribute to the choice of primary treatment.15
Since the Institute of Medicine's 1996 Health Care Quality Initiative, interest in the science of quality of care assessment has increased. Although geographic variation itself does not constitute poor-quality care, this variation raises concern. Variation based on clinical parameters may indicate uncertainty in standard practice guidelines; variation based on sociodemographic characteristics indicates a differential utilization of available resources and, potentially, poor quality of care. Determining whether geographic, ethnic, and socioeconomic variations exist in the treatment of men with localized prostate cancer provides a foundation for improving quality of care for the large number of men facing this disease.
In this study, we set out to define target areas for quality improvement by assessing the influence of nonclinical factors on primary treatment rates. We used the Surveillance, Epidemiology, and End Results (SEER) public use files to identify a national sample of patients with early-stage prostate cancer, including those younger than 65 years of age who lack Medicare coverage. We hypothesized that in the PSA era, primary treatment rates would have become established locally but not nationally, and that socioeconomic factors would independently influence treatment rates.
METHODS
Data Sources
SEER is a population-based tumor registry maintained by the National Cancer Institute for 11 regions in the United States chosen to represent epidemiologically significant population subgroups.16 The 11 regions comprise five states (Connecticut, Hawaii, Iowa, New Mexico, and Utah) and six metropolitan areas (Detroit, San Francisco, Atlanta, Seattle, Los Angeles County, and San Jose-Monterey). These regions reflect the general population in terms of education and poverty level, although for privacy purposes these variables are not recorded for each individual.
Study Population
We identified all patients older than age 20 diagnosed between 1995 and 1999 with a stage of localized/regional prostate cancer. Beginning in 1995, SEER collapsed all prostate cancer patient categories defined as "confined entirely to the organ" or "extended beyond the limits of the organ of origin directly into surrounding tissue or lymph nodes" into the single category of localized/regional. For each patient, we identified the following variables: county of residence, age at diagnosis, race/ethnicity, date of diagnosis, tumor grade, site-specific surgery, radiation, and radiation sequence with surgery.
We defined primary treatment as surgery (coded as radical/total prostatectomy), radiation, or watchful waiting. Because our focus was on primary treatment, we classified patients who underwent radiation subsequent to radical prostatectomy only as surgery. For this analysis, radiation therapy included beam radiation, radioactive implants, or a combination of the two methods. To be classified as watchful waiting, we required both a surgical code of "no surgery" and a radiation code of "none." patients with treatment coded as "recommended, unknown if done" or "unknown" were excluded from analysis. Patients were also excluded if they had radiation before surgery (due to incomplete radiation data), incomplete ethnicity or grade information, or insufficient data to determine the treatment sequence.
We created two socioeconomic variables (income and education) using data from the year 2000 US Census. The US Census Bureau Web site provides a tool called the American FactFinder that allows the user to create specific queries from the Census 2000 summary file. We created custom tables identifying the median educational attainment and median annual household income in 1999 dollars for men older than 25 years. We further stratified men by specific ethnicity (black alone, American Indian/Alaskan Native, Asian alone, Hispanic, white alone not Hispanic, and other) residing in each of the corresponding SEER counties. The US Census counties matched the SEER code for county precisely. We ascribed the median education and income level from the census data to each individual analyzed from the SEER data based on his county of residence and ethnicity.
Analysis
Descriptive statistics are presented for demographic and clinical data. Multiple comparisons of primary treatment rates in each region referenced against the composite rate were compared using a Bonferroni correction with an overall error rate of 0.05. Similar comparisons were assessed at the intraregional level. We performed multinomial logistic regression analysis with watchful waiting as the referent group to investigate a link between the likelihood of primary treatment and demographic variables. In addition to age, we included geographic location and grade as covariates. Age remained a continuous variable in the model, whereas income was dichotomized at the lowest quartile of $40,216. We categorized ethnicity as white, black, Hispanic, and other; we collapsed education level into less than high school, high school graduate, and college educated. We categorized tumor grade as poorly, moderately, or well-differentiated, and specified Los Angeles as the referent SEER region. All analyses were conducted with SAS software, version 8 (SAS Institute, Cary, NC).
RESULTS
During 1995 to 1999, 102,970 patients had stage coded as localized/regional prostate cancer and coding that allowed clear delineation of treatment type. Of these, we could not ascertain ethnicity and grade in 2,118 and 4,234, respectively. Many patients had multiple missing variables, leaving 96,769 (94%) men for inclusion in the final multivariate analysis. We compared the 6,201 patients omitted from the final analysis to the original sample and did not find any statistical differences.
Table 1 lists patients' demographic and clinical characteristics, stratified by treatment modality. Patients undergoing surgery were significantly younger, were more likely to be white, and tended to have less than a high school education. Compared with those undergoing surgery, radiation patients tended to have a higher income and be older; those undergoing watchful waiting were also significantly older and had a greater likelihood of well-differentiated tumors. Patients in Los Angeles underwent surgery or watchful waiting at significantly higher rates than did patients in other regions, whereas those in Detroit had notably higher radiation rates.
The patterns reflecting these significant geographic variations are shown in Fig 1. Because of the large sample size, we were able to detect small differences between regional and composite proportions. Of the regional surgery proportions, only New Mexico and Seattle did not differ statistically from the composite proportion. Among the remaining nine regions, five (Atlanta, Iowa, Los Angeles, San Jose, and Utah) had proportions above the composite, whereas four (Connecticut, Hawaii, Detroit, and San Francisco) had proportions below the composite. All of the regional proportions for radiation differed significantly from the composite radiation proportion. Radiation proportions were higher than the composite in Connecticut, Detroit, San Francisco, Atlanta, and Seattle. The composite watchful waiting proportion differed significantly from the regional proportions in Atlanta, Connecticut, Detroit, New Mexico, San Jose, Seattle, and Utah. Connecticut, New Mexico, and Utah had higher watchful waiting proportions than either the composite proportion or the other regions. These differences persisted when patients were stratified by age to younger or older than 70.
Within-region analyses (data not shown) revealed more subtle variations. For all 11 regions, treatment rates were much more consistent intraregionally than inter-regionally. For example, within San Jose, only one county differed statistically from the regional average for watchful waiting and radiation, and none differed for surgery.
We constructed multivariate models to assess the association of education, income, and tumor grade with primary treatment rates, after adjusting for region (Table 2). The odds of blacks undergoing surgery rather than watchful waiting were half those of whites. The odds of lower income patients undergoing surgery were 20% less than for higher income patients. Tumor grade was the strongest predictor of surgical therapy: the less differentiated the tumor, the more likely the patient was to undergo surgery.
Similarly, the odds of blacks receiving radiation therapy instead of watchful waiting were 20% lower than for whites. Other ethnicity, which includes men of American Indian/Alaska Native, Chinese, Filipino, Japanese, or Hawaiian ancestry, was also associated with lower radiation rates when compared with watchful waiting rates in whites. The odds of lower income patients receiving radiation treatment were one fourth lower than for whites. In summary, these patients (lower income, blacks, and other ethnicities) had markedly lower rates of radiation therapy. Lastly, higher tumor grade predicted radiation over watchful waiting.
DISCUSSION
Widespread interest in measuring quality has emerged, as consumers and third-party payers increasingly try to identify high-quality providers and institutions.17,18 However, the determination of high-quality care is often subjective. In his model of quality assessment, Donabedian2 proposed measuring structure as available medical resources, process as technical aspects of care, and outcomes as patients' ultimate health status. This paradigm has become the basis for contemporary health care quality assessment. As investigators move toward developing process measures to improve quality of care, highly prevalent conditions are being targeted.19 We examined geographic, ethnic, and socioeconomic variation in the treatment of men with prostate cancer, a highly prevalent disease, and identified three critical findings.
First, in the 10 years since the original geographic variation studies in prostate cancer, there has been minimal convergence in treatment rates at the national level. Prior studies found the highest surgical rates in the west and the lowest rates in the northeast—a pattern we continue to see.6,20,21 Our findings likewise corroborate prior studies of the use of radiation therapy.7 Age has been implicated as a mitigating factor in previous investigations of variation in radiation or surgical treatment rates; however, we found variations after controlling for age.7,20 Although current research suggests clinical equipoise among prostate cancer treatments, providers appear to favor the modality associated with their specialty. For example, urologists tend to trust the results of radical retropubic prostatectomy over radiation therapy, perhaps because the long natural history of prostate cancer creates difficulties assessing the long-term cancer control of the more recent radiation therapies. In addition, our multivariate analysis suggests that even among radiation oncologists, geographic variation exists in practice patterns. The odds ratios for regional variation were two-fold higher in Hawaii and Atlanta, compared with Connecticut and Los Angeles. We found less surgical geographic variation, with fewer regions demonstrating significant variation, and those differing did so to a lesser degree.
At the intraregional level, treatment rates are much more uniform. This contrasts with other studies that demonstrated more uniform rates at the national level. These investigators hypothesized that larger areas were better able to distribute the variance uniformly. Thus, smaller regions demonstrated more variation.22,23 Perhaps in the PSA era, earlier diagnosis has produced a more clinically homogeneous patient population, leading to diminished variation. However, a Connecticut study of prostate cancer incidence during the PSA era found widespread variation in incidence rates across even small geographic areas. The authors postulated that the variation reflected differential screening practices by individual practitioners, and not differences in the underlying disease burden in the population.24 Similarly, we doubt that there are underlying differences in disease burden by region. Unlike the Connecticut study, our analysis found uniform prostate cancer treatment rates at the intraregional level, suggesting that local consensus may exist in the management of men with early-stage prostate cancer.
Although geographic variation originally was studied in terms of resource consumption, today the implications of geographic variation extend to quality of care.25 Geographic variation does not necessarily translate to poor quality of care. Because the competing modalities appear equally efficacious for most patients, treatment differences may be a result of available resources and not individual preference. However, an examination of Veterans Affairs Medical Centers in the contiguous United States found a similar pattern of geographic variation in surgical rates,26 suggesting that surgical variation persists even in the face of adequate resources. In fact, Ashton et al27 found that hospital resources, defined as number of beds, actually correlated with use. This pattern was consistent for eight different chronic diseases. In this era of evidence-based medicine, such variations speak to the need to determine which patients are best served by which treatment modalities. Trials such as the recently abandoned American College of Surgeons Oncology Group Z0070, a randomized, controlled trial comparing surgery with brachytherapy, are critical to the elucidation of this issue.
Second, ethnicity may be associated with treatment rates. Ethnicity and SES must be examined simultaneously because of the high prevalence of poverty among ethnic minorities. Intuitively, different cultural backgrounds could predispose patients to certain treatment preferences.28 Several studies have documented higher radiation rates in blacks, and focus group work in this population confirms a cultural bias toward "avoiding the knife."13,21,29
However, receiving radiation as treatment is different than receiving no treatment at all. Our data showed that blacks of comparable stage as that of whites or Hispanics were more likely to receive no treatment when compared with surgery or radiation therapy. This may suggest a lack of education regarding the natural history of prostate cancer and available treatment options. There is a tendency toward conservative treatment in blacks in the Medicare population, in whom more advanced age, comorbidity, and the indolent nature of prostate cancer may heavily influence the decision-making process.15 The role of education in the treatment decision is speculative; social support, access to care, transportation issues, or physician bias are other potential factors influencing the decision. The crucial question is whether the patients are involved in that decision-making process and actively choose watchful waiting or are steered on that course by their physician. Furthermore, watchful waiting suggests monitoring of a known condition. Shavers et al30 found that blacks and Hispanics were less likely than whites to receive active monitoring during a 60-month period, suggesting ethnic disparities in prostate cancer management.30 A more qualitative examination of patient and physician decision making would allow for exploration of these issues. Quality is not impaired but rather is preserved if patient participation and respect for cultural beliefs drive the ethnic disparity. However, it behooves researchers to undertake analyses providing evidence for these assertions.
Third, we found an association between low income and watchful waiting, independent of age, ethnicity, disease stage, and geographic region. Although ethnic minorities tend to occupy a lower socioeconomic position, blacks and those of low income are predisposed to watchful waiting. Analysis of the Connecticut Tumor Registry supports the concept that poverty inversely predicts radical prostatectomy rates.31 In fact, when stage and poverty level were controlled for, ethnicity was no longer a significant predictor of treatment. The Connecticut Tumor Registry included a poverty variable at the census tract level. Klabunde et al13 confirmed that men residing in higher poverty regions, as determined by census tract, had lower surgery rates. These studies provide external validity for our use of a separate area-based measure of SES, namely the median income at the county level. We recognize that all of these measures are proxies for individual SES; however, because public health surveillance systems have not routinely collected individual data, researchers must use such proxies until such data become available.
Tarman et al14 used military rank as a surrogate for SES and found that patients of lower SES had significantly higher Gleason grade at presentation; thus, low-income patients are potentially most at risk from their disease. We found that the strongest predictor of surgery or radiation therapy was a moderate- or high-grade tumor, yet the low-income patients in our study more often received no treatment, after adjusting for grade.
Although those of lower income may have lacked adequate access to health care, they must have had access to some portion of the healthcare system to be included in the tumor registry. From our data set, it is unclear if the treatment disparity is due to poor quality of care within the health plan or access to a lower quality health plan. Analyses of a variety of medical activities including breast cancer screening, eye examinations in diabetics, knee arthroplasty, and the use of beta-blockers have found that within health plans, patients of different ethnicities are treated equally. However, socioeconomic variation persisted across samples because lower income patients were more commonly enrolled in lower quality health plans.5,32
The finding that lower income predisposes patients to watchful waiting as a therapeutic option has implications for quality of care, regardless of whether it occurs at the provider or health plan level. If the regional availability of medical resources influences the decision, it is hard to fathom why this would preferentially affect low-income patients. Carlisle et al4 demonstrated that ethnicity and income are more important determinants of surgical treatment rates than the availability of medical services. This may be the more concerning quality problem to combat. Provider-propagated differences may be implicated by our finding that education level was not associated with treatment rates. Less-educated patients may have a worse understanding of the nuances of treatment and difficulty comprehending the available literature, thus they might make less-informed choices.33,34 However, education did not affect treatment rates in our patients. Conversely, low-income patients may have comorbid conditions that preclude providers from offering them aggressive therapy. Another potential reason for the disparity in treatment options beyond the provider's control relates to hidden economic costs. Low-income patients may opt for less aggressive treatment because they perceive that they cannot afford time off from work or they have transportation issues.
Our study has several limitations. First, we had to ascribe group characteristics to individuals, which may not accurately reflect the income or education of that individual. However, given that public health surveillance systems lack individualized data, proxies must be used. Leading researchers in geocoding have found that area-based geosocial measures can be applied validly to individuals residing in that area.35,36 Many prostate cancer researchers have used SEER-Medicare–linked databases, which provide socioeconomic data at the census tract level. Although there are advantages to this approach, it fails to identify treatment patterns in younger men diagnosed with prostate cancer. Our analysis is particularly robust because it included men of all ages. Those younger than Medicare age are increasingly being diagnosed as PSA screening becomes more ubiquitous. Second, we did not have a detailed assessment of patients' health status. Although we hypothesized that men younger than 70 years of age with local/regional disease are comparable, this may not be the case. Comorbid conditions often drive therapy considerations in prostate cancer patients. Third, no data on patient preference or provider characteristics are available through SEER, and this could be an important underlying reason for the variations we observed. The characteristics of physicians caring for the lower income patients may differ from those of other physicians in terms of board certification, fellowship training, or practice duration, and may influence quality of care. Fourth, SEER limits data collection to 4 months from initial diagnosis. The inefficiencies of health systems serving the disadvantaged could lead to treatment delays extending beyond 4 months. Therefore, our data for the men of lower SES may not accurately reflect treatment received. However, this delay itself may negatively influence quality; researchers from Canada have found that a delay of 3 months may adversely affect cancer control.37 Lastly, we infer that different treatment rates suggest inequities in care. These treatment differences may be just. The grouping of localized disease with regional disease potentially confounds this issue because low-income patients are more likely to present with more advanced disease. This would predispose them to radiation, but not watchful waiting. Clinical parameters or underlying health status, not SES or ethnicity, should be the reason for the inequality.
Geographic, ethnic, and socioeconomic variations in prostate cancer treatment persisted into the late 1990s. These findings have implications for quality of care in that they may suggest unconscious discrimination on the part of physicians. Systematic studies of outcomes in this population, controlling for the confounding variables of ethnicity and SES, will help answer these questions. With better evidence to identify which patient populations benefit from each treatment modality, we will be able to disseminate these findings nationwide to improve quality of care for men with prostate cancer.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported by a grant from the American Foundation for Urologic Disease/American Urological Association Research Scholar Program.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Christianson R: ACHA 1998 Quality Award winners: Quality initiatives that stand out. Provider 24:81-89, 1998
Donabedian A: Quality assurance: Structure, process and outcome. Nurs Stand 7:4-5, 1992
Wennberg J, Gittelsohn A: Variations in medical care among small areas. Sci Am 246:120-134, 1982
Carlisle DM, Valdez RB, Shapiro MF, et al: Geographic variation in rates of selected surgical procedures within Los Angeles County. Health Serv Res 30:27-42, 1995
Skinner J, Weinstein JN, Sporer SM, et al: Racial, ethnic, and geographic disparities in rates of knee arthroplasty among Medicare patients. N Engl J Med 349:1350-1359, 2003
Lu-Yao GL, McLerran D, Wasson J, et al: An assessment of radical prostatectomy: Time trends, geographic variation, and outcomes—The Prostate Patient Outcomes Research Team. JAMA 269:2633-2636, 1993
Lai S, Lai H, Lamm S, et al: Radiation therapy in non-surgically-treated nonmetastatic prostate cancer: Geographic and demographic variation. Urology 57:510-517, 2001
Smith RP, Malkowicz SB, Whittington R, et al: Identification of clinically significant prostate cancer by prostate-specific antigen screening. Arch Intern Med 164:1227-1230, 2004
Hugosson J, Aus G, Lilja H, et al: Results of a randomized, population-based study of biennial screening using serum prostate-specific antigen measurement to detect prostate carcinoma. Cancer 100:1397-1405, 2004
Sylvester JE, Blasko JC, Grimm PD, et al: Ten-year biochemical relapse-free survival after external beam radiation and brachytherapy for localized prostate cancer: The Seattle experience. Int J Radiat Oncol Biol Phys 57:944-952, 2003
Pound CR, Partin AW, Eisenberger MA, et al: Natural history of progression after PSA elevation following radical prostatectomy. JAMA 281:1591-1597, 1999
D'Amico AV, Whittington R, Malkowicz SB, et al: Biochemical outcome after radical prostatectomy or external beam radiation therapy for patients with clinically localized prostate carcinoma in the prostate specific antigen era. Cancer 95:281-286, 2002
Klabunde CN, Potosky AL, Harlan LC, et al: Trends and black/white differences in treatment for nonmetastatic prostate cancer. Med Care 36:1337-1348, 1998
Tarman GJ, Kane CJ, Moul JW, et al: Impact of socioeconomic status and race on clinical parameters of patients undergoing radical prostatectomy in an equal access health care system. Urology 56:1016-1020, 2000
Shavers VL, Brown ML, Potosky AL, et al: Race/ethnicity and the receipt of watchful waiting for the initial management of prostate cancer. J Gen Intern Med 19:146-155, 2004
Surveillance, Epidemiology, and End Results (SEER) Program Public-Use Data (1973-2001): National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, 2003, based on November 2002 submission. http://www.seer.cancer.gov
Brook RH, McGlynn EA, Shekelle PG: Defining and measuring quality of care: A perspective from US researchers. Int J Qual Health Care 12:281-295, 2000
Birkmeyer JD, Dimick JB: Potential benefits of the new Leapfrog standards: Effect of process and outcomes measures. Surgery 135:569-575, 2004
Spencer BA, Steinberg M, Malin J, et al: Quality-of-care indicators for early-stage prostate cancer. J Clin Oncol 21:1928-1936, 2003
Lai S, Lai H, Krongrad A, et al: Radical prostatectomy: Geographic and demographic variation. Urology 56:108-115, 2000
Harlan L, Brawley O, Pommerenke F, et al: Geographic, age, and racial variation in the treatment of local/regional carcinoma of the prostate. J Clin Oncol 13:93-100, 1995
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 307:1310-1314, 1982
McPherson K, Strong PM, Epstein A, et al: Regional variations in the use of common surgical procedures: Within and between England and Wales, Canada and the United States of America. Soc Sci Med [A] 15:273-288, 1981
Gregorio DI, Kulldorff M, Sheehan TJ, et al: Geographic distribution of prostate cancer incidence in the era of PSA testing, Connecticut, 1984 to 1998. Urology 63:78-82, 2004
Wennberg JE, Freeman JL, Culp WJ: Are hospital services rationed in New Haven or over-utilised in Boston? Lancet 1:1185-1189, 1987
Wilt TJ, Cowper DC, Gammack JK, et al: An evaluation of radical prostatectomy at Veterans Affairs Medical Centers: Time trends and geographic variation in utilization and outcomes. Med Care 37:1046-1056, 1999
Ashton CM, Petersen NJ, Souchek J, et al: Geographic variations in utilization rates in Veterans Affairs hospitals and clinics. N Engl J Med 340:32-39, 1999
Knight SJ, Siston AK, Chmiel JS, et al: Ethnic variation in localized prostate cancer: A pilot study of preferences, optimism, and quality of life among black and white veterans. Clin Prostate Cancer 3:31-37, 2004
Sohayda CJ, Kupelian PA, Altsman KA, et al: Race as an independent predictor of outcome after treatment for localized prostate cancer. J Urol 162:1331-1336, 1999
Shavers VL, Brown M, Klabunde CN, et al: Race/ethnicity and the intensity of medical monitoring under ‘watchful waiting’ for prostate cancer. Med Care 42:239-250, 2004
Polednak AP: Prostate cancer treatment in black and white men: The need to consider both stage at diagnosis and socioeconomic status. J Natl Med Assoc 90:101-104, 1998
Schneider EC, Leape LL, Weissman JS, et al: Racial differences in cardiac revascularization rates: Does "overuse" explain higher rates among white patients? Ann Intern Med 135:328-337, 2001
Kane CJ, Lubeck DP, Knight SJ, et al: Impact of patient educational level on treatment for patients with prostate cancer: Data from CaPSURE. Urology 62:1035-1039, 2003
Kim SP, Knight SJ, Tomori C, et al: Health literacy and shared decision making for prostate cancer patients with low socioeconomic status. Cancer Invest 19:684-691, 2001
Krieger N, Williams DR, Moss NE: Measuring social class in US public health research: Concepts, methodologies, and guidelines. Annu Rev Public Health 18:341-378, 1997
Krieger N: Overcoming the absence of socioeconomic data in medical records: Validation and application of a census-based methodology. Am J Public Health 82:703-710, 1992
Nam RK, Jewett MA, Krahn MD, et al: Delay in surgical therapy for clinically localized prostate cancer and biochemical recurrence after radical prostatectomy. Can J Urol 10:1891-1898, 2003(Tracey L. Krupski, Lorna )