Uterine cancer is the most common gynecologic malignancy in the United States with approximately 47,130 new diagnoses and 8010 deaths in 2012.1 Surgical staging of uterine cancer is recommended to more accurately delineate the extent of disease.2 When a gynecologic oncologist performs this surgical staging, patients with early-stage disease are less likely to receive unnecessary radiation therapy, and patients with advanced-stage disease are more likely to receive adjuvant chemotherapy.3,4
In addition to access to specialty care, access to high-volume cancer care seems to improve treatment and survival. Hillner et al5 performed a meta-analysis evaluating the relationship between high-volume hospitals (HVHs) and surgeons (HVSs) and cancer patient outcome. Most investigations reviewed in this meta-analysis demonstrated a positive relationship between high-volume cancer care and cancer patient outcome. Access to high-volume care specifically for patients with gynecologic cancer has also been evaluated. Based on Maryland-specific and national databases, access of patients with primary ovarian cancer to HVHs and HVSs is associated with a higher likelihood of receiving standard therapy and lower risk of in-hospital death.6,7 However, despite the possible negative impact, low-volume surgeons at low-volume hospitals (LVHs) perform a substantial number of primary ovarian cancer surgeries in Maryland.8 Similar to ovarian cancer, uterine cancer surgery by an HVS has been associated with a 48% reduction in risk of in-hospital death.9 Additionally, based on data by Fleury et al,10 race does not seem to be associated with in-hospital mortality or surgery at an HVH for women undergoing uterine cancer surgery.
One barrier to high-volume care seems to be distance. Access of patient with ovarian cancer to HVSs in Maryland was negatively associated with residence 50 miles or more from an HVH.8 A statewide analysis of access to surgical care of endometrial cancer in Arizona by Benjamin et al11 demonstrated that 61% of endometrial cancer cases were performed by HVSs, and one fifth of patients traveled 50 miles or more to receive care. Individuals seeking care at an HVH were more likely to travel 50 miles or more to receive care. These investigations illustrate 2 ways in which distance can be associated with access to high-volume care. Patients who live remote from a high-volume center are less likely to obtain care at an HVH. However, patients who travel a substantial distance to their treating hospital are more likely to do so to obtain care at an HVH. To date, no investigation has evaluated the influence of both factors in the same study population. The primary study objective was to evaluate the influence of both distance from treating hospital and distance to nearest HVH on access to high-volume surgical care of uterine cancer in Maryland.
MATERIALS AND METHODS
The Institutional Review Board of Johns Hopkins Hospital approved this investigation (NA 00035327) involving a retrospective cross-sectional analysis of hospital discharge data from nonfederal acute care hospitals in Maryland. The Maryland Health Services Cost Review Commission (HSCRC) is an independent state agency that discloses hospital data to establish rates for payers. The HSCRC database contains information related to all procedures performed statewide with associated patients’ demographics, treating hospital name, hospital costs incurred, and 30 days of follow-up data but does not contain information on stage, grade, or long-term follow-up. The HSCRC database was searched using International Classification of Disease, Ninth Revision codes to identify all women 18 years or older who were admitted for management of all histologic types of uterine cancer from January 1, 1994, to December 31, 2010. Current Procedural Terminology codes for hysterectomy, bilateral salpingo-oophorectomy, and lymphadenectomy were searched to identify patients undergoing primary surgical staging of uterine cancer.
Relevant clinical data including race, age, marital status, patient’s residence zip code, insurance type, year of surgery, name of treating hospital, and the unique identifier of the treating surgeon was extracted from the HSCRC database. Zip codes for treating hospitals were procured by Internet search for and evaluation of each hospital’s Web site. Average annual surgical volumes of surgeon and hospital were calculated. Total number of cases was divided by the number of years in practice in Maryland (for surgeons) or in operation (for hospitals). Based on previous definitions by Bristow et al,8 an average of 10 or more and 20 or more annual primary uterine cancer surgeries characterized an HVS and an HVH, respectively. United States Census designations were accessed to classify the community setting of home zip codes as ”rural“ (<2500 persons), ”suburban“ (2500–50,000 persons), or ”urban“ (>50,000 persons).12
Distance to hospital was calculated by first collecting latitude and longitude coordinates for each zip code. Using coordinate pairings, a trigonometric formula was applied to calculate the distance between home and hospital zip codes. These calculations represent the shortest possible distance between 2 points and thus do not denote actual driving distances. Although the earth is spheroid, for the relatively short travel in this study, these calculations closely approximate ”as the crow flies“ distances. Based on previous definitions by Benjamin et al11 and Bristow et al,8 we defined distance to treating hospital as local (<50 miles) or distant (≥50 miles), and we defined distance to the nearest HVH as local (<50 miles) or distant (≥50 miles).
Univariable and multivariable logistic regression analyses were performed with likelihood of surgery at HVH, distance 50 miles or more to treating hospital, or distance 50 miles or more to the nearest HVH as the outcome variable. Race, age, HVS, time period of surgery, rural location, and payer status were covariates. Before modeling for multivariable logistic regression, individual factors were evaluated for correlation. Pearson correlation coefficients greater than 0.55 were considered sufficient evidence to remove one of the factors from the model. All statistical analyses were performed using Stata 11.1 statistical software (StataCorp, College Station, TX).
From 1994 to 2010, 9671 patients with uterine cancer were admitted to nonfederal Maryland hospitals. 9585 women had valid geographic information. Of these, 8529 women (88%) underwent primary surgical management of uterine cancer. Procedures were performed by 849 surgeons at 47 hospitals in Maryland. Of the 8529 patients, 54.9% (4682) were younger than 65 years. Whites, African Americans, and Asians accounted for 77.8% (6636/8529), 17.4% (1480/8529), and 1.3% (109/8529) of the population, respectively. Of the 8529 patients, 47.2% (4027) had government insurance (Medicare/Medicaid). Most (82.2% [7012/8529]) of the patients had a suburban zip code. Of the 8529 patients, 46.4% (3956) had surgery from 1994 to 2002, and 53.6% (4573/8529) had surgery between 2003 and 2010. Of the 8529 patients, 9.2% (783) traveled 50 miles or more to their treating hospital; 17.2% (1466/8529) of the women resided 50 miles or more from the nearest HVH. Only 3.3% (28/849) of surgeons in Maryland were classified as high volume. These 28 HVS performed 50% (4281/8529) of the uterine cancer cases. Of 47 hospitals, 14.9% (7) were considered HVH. Most (60.8% [5189/8529]) of the procedures were performed at HVHs (Table 1).
Surgery at HVH
Univariable analysis demonstrated white race, married status, distance 50 miles or more from nearest HVH, and rural zip code to have a significantly lower likelihood of having primary uterine cancer surgery at an HVH (married, P = 0.017; otherwise, P < 0.001) Travel 50 miles or more to the treating hospital, surgery by an HVS, and surgery during the latter half (2003–2010) of the investigation were each associated with a significantly higher likelihood of having primary uterine cancer surgery at an HVH (all, P < 0.001). Age and insurance type were not significant factors (Table 2).
Multivariable analysis was performed. A significant correlation was noted between distance to treating hospital and distance from nearest HVH (Pearson correlation coefficient, 0.581; P < 0.001). Therefore, we performed 2 separate multivariable logistics regression analyses using likelihood of undergoing surgery at an HVH as the outcome variable. The first analysis used distance to treating hospital as a covariate and demonstrated white race and rural zip code to have a significantly lower likelihood of having surgery at an HVH (both, P < 0.001) Travel 50 miles or more to treating hospital, surgery by an HVS, and surgery during the latter half (2003–2010) of the investigation were associated with a significantly higher likelihood of having surgery at an HVH. (all, P < 0.001) The second analysis used distance to the nearest HVH as a covariate. This analysis demonstrated similar findings except that distance 50 miles or more to the nearest HVH was associated with a significantly lower likelihood of having surgery at an HVH (P < 0.001; Table 2).
Travel 50 Miles or More to Treating Hospital
Patients who travel 50 miles or more to their treating hospital were more likely to undergo surgery for primary uterine cancer at an HVH. Univariable analysis demonstrated age 65 years or older, surgery at an HVH, distance from the nearest HVH 50 miles or more, rural zip code, surgery by an HVS, and surgery during the latter half (2003–2010) of the investigation to be associated with a significantly higher likelihood of travel 50 miles or more to the treating hospital (age 65 years or older, P = 0.002; otherwise, P < 0.001).
Given the correlation with travel to the treating hospital, we did not use distance from the nearest HVH in the multivariable analysis. To improve comparisons with the other outcome variables, we used the covariates race, marital status, location, and surgeon volume in addition to hospital volume. White race, surgery at an HVH, rural zip code, and surgery by an HVS were associated with a significantly higher likelihood of traveling 50 miles or more to the treating hospital (all, P < 0.001; Table 3).
Distance From Nearest HVH 50 Miles or More
Patients who resided 50 miles or more from the nearest HVH were less likely to undergo surgery for primary uterine cancer at an HVH. Univariable analysis demonstrated surgery at an HVH (P < 0.001) and surgery by an HVS (P = 0.026) to be associated with a significantly lower likelihood of residence 50 miles or more from an HVH. White race, married status, distance 50 miles or more to the treating hospital, rural zip code, and government insurance were associated with a significantly higher likelihood of residence 50 miles or more from an HVH (government insurance, P = 0.034; otherwise, P < 0.001).
Given the correlation with distance from the nearest HVH, we did not use distance traveled to the treating hospital in the multivariable analysis. To improve comparisons with the other outcome variables, we used the covariates race, marital status, location, and surgeon volume in addition to hospital volume. Surgery at an HVH was associated with a significantly lower likelihood of residing 50 miles or more from an HVH (P < 0.001). White race, rural zip code, surgery by an HVS, and surgery during the latter half (2003–2010) of the investigation were associated with a significantly higher likelihood of residing 50 miles or more from the nearest HVH (later half of investigation, P = 0.004; otherwise, P < 0.001; Table 4).
Maryland Geography and Racial Distribution
We evaluated Maryland population distribution by county and racial composition and by the location of HVH. Of the 24 Maryland counties, 5 counties contain 66% of the population. Whites represent most of the population in 3 of these counties (Montgomery, Baltimore County, and Anne Arundel). African Americans comprise the majority in the other 2 counties (Prince George’s and Baltimore City). Baltimore City, Baltimore County, and Montgomery County have 4, 2, and 1 HVHs, respectively. The 2 other high-population counties (Prince George and Anne Arundel) are immediately adjacent to counties possessing HVH. The presence of 4 HVHs within Baltimore City is associated with higher fraction of primary uterine cancer surgeries than its population would suggest. Although only 12.3% (1052/8529) of patients with primary uterine cancer reside in Baltimore City, 38.5% (3285/8529) of primary uterine cancer surgeries were performed in Baltimore City (Table 5).12
Most (82.8% [7063/8529]) of the patients with uterine cancer in our investigation reside less than 50 miles from the nearest HVH. Fifty-mile radius circles were created around each the 7 HVHs. Only 2 western counties and 3 Eastern Shore counties are entirely outside the 50-mile radius of all HVHs. Less than 5% of the Maryland population live in these 5 counties (Fig. 1).
Interestingly, 7.0% (595/8529) of patients with primary uterine cancer surgery were treated at an LVH when an HVH was closer. Referral bias may play a role, as 33% (195/595) of these patients had their surgery performed by an HVS. As traffic patterns may differ from ”as the crow flies“ distances, calculations were limited to patients who traveled substantially farther than the closest HVH. Only 2% (170/8529) and 0.5% (45/8529) of patients with primary uterine cancer surgery were treated at an LVH when an HVH was 5 miles and 15 miles closer, respectively.
In Maryland, white race, rural residence, and residence 50 miles or more from the nearest HVH are associated with a significantly decreased likelihood of undergoing primary uterine cancer surgery at a HVH. Surgery by an HVS, surgery from 2003 to 2010, and travel 50 miles or more to the treating hospital were associated with significantly increased likelihood of having primary uterine cancer surgery at an HVH. Furthermore, white race, rural residence, surgery by an HVS, and surgery from 2003 to 2010 were all associated with a significantly increased likelihood of both residence 50 miles or more from the nearest HVH as well as travel 50 miles or more to the treating hospital. These 2 measures of distance have opposing effects on location of care. Travel 50 miles or more to the treating hospital was associated with a significantly increased likelihood of surgery at an HVH. However, residence 50 miles or more from the nearest HVH was associated with a significantly decreased likelihood of having surgery at an HVH.
Our group’s earlier evaluation of this data set from 2000 to 2009 showed a significant association between African American race and surgery by an HVS but not surgery at an HVH.10 Two issues may explain these findings. First, we evaluated patients over a longer study period (1994–2010), which would increase our statistical power to detect a difference. In addition, rather than plotting the data to determine a breakpoint for the number of cases per year, we used definitions reported by Bristow et al8 and used by Benjamin et al.11
In Maryland, white race and rural residence are associated with a lower likelihood of surgery at an HVH. With regard to race, our findings contrast those seen in Arizona and may be explained by differences in population distribution. In Maryland, nearly half of the white population lives outside the most populated counties, which have the closest proximity to high-volume care. Therefore, it is not unexpected that white patients are more likely to reside 50 miles or more from an HVH and more likely to travel 50 miles or more to receive care. In contrast, the most of the whites in Arizona reside in the most populous counties, and these patients are more likely to have surgery at an HVH. Most Native Americans in Arizona live outside the most populous counties, which may explain the lower likelihood of surgery at an HVH.11 Our data suggest that when nonwhite populations live in reasonable proximity to high-volume centers, they are likely to use these facilities.
In Arizona, from 2005 to 2008, most patients with uterine cancer had surgery at an HVH (68%) by an HVS (61%).11 Overall, in Maryland from 1994 to 2010, a smaller proportion of patients with uterine cancer underwent surgery at an HVH (61%) by an HVS (50%). However, during the latter half of the study (2003–2010), these numbers improved with percentages nearly identical to Arizona. These changes may reflect a better understanding over time of both patients and primary providers as to the importance of cancer care at high-volume centers.
There are several limitations to our study. The distance calculation we performed does not reflect actual driving distances but rather provides an estimate based on the shortest distance between residence and hospital. The lack of a universal Current Procedural Terminology code for primary uterine cancer surgery may have prevented proper inclusion or exclusion. The HSCRC database does not include information regarding the International Federation of Gynecology and Obstetrics stage, grade, histology, and long-term follow-up. Additionally, the database does not collect data on socioeconomic issues such as income, education, and occupation. The data were evaluated retrospectively. Race was either self-reported or interpreted by hospital personnel, which may lead to classification bias. Comparing patients over a 17-year period lends to variation in the management of patients with uterine cancer. However, it also provides a substantial study population. Lastly, since we evaluated only patients undergoing primary surgical management of uterine cancer, disparities associated with the management of inoperable or advanced uterine cancer may not be reflected in this investigation.
To the best of our knowledge, our investigation is the first to evaluate the impact of both distance to the nearest HVH and distance traveled to treating hospital on access to care for a single cancer in a single state. In Maryland, patient residence 50 miles or more to the nearest HVH creates a significant barrier to surgical management of uterine cancer at a high-volume center. However, patients who travel 50 miles or more are more likely to receive care by an HVS at an HVH. Nonwhites who live in proximity to HVH are likely to use these services. Further research is needed to identify mechanisms to mitigate the obstacle that distance imposes on access to high-volume care. As some patients seem willing to travel substantial distances, education of primary providers on the importance of high-volume cancer care may be beneficial. However, many patients may be unwilling or unable to travel to high-volume centers unless transportation is provided or these services are available locally.
The authors thank Kymberlee Olson, clinical analytics systems architect, for assistance with the HSCRC database.