Predictors of comprehensive surgical treatment in patients with ovarian cancer


  • The CDC investigator was involved in the study conceptualization, analysis plan, interpretation of the data, and review of the article. The article underwent standard CDC clearance (review) procedures and was approved for publication.

  • The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the CDC.



Providing appropriate surgical treatment for women with ovarian cancer is one of the most effective ways to improve ovarian cancer outcomes. In this study, the authors identified factors that were associated with a measure of comprehensive surgery, so that interventions may be targeted appropriately to improve surgical care.


Using Healthcare Cost and Utilization Project hospital discharge data from 1999 to 2002 for 9 states, the authors identified 10,432 admissions of women who had an International Classification of Disease, 9th Revision (ICD-9) primary diagnosis of ovarian cancer and who had undergone oophorectomy. Based on National Institutes of Health Consensus Panel recommendations, surgeries were categorized as comprehensive by using ICD-9 diagnosis and procedure codes. Logistic regression analysis using data from 5 states with a full set of variables (n = 6854 patients)was used to identify factors that were associated with the receipt of comprehensive surgical care.


Overall, 66.9% of admissions (range, 46.3–80.8% of admissions) received comprehensive surgery. Factors that were associated independently with comprehensive surgical care included age (ages 21–50 years vs ages 71–80 years or ≥81 years), race (Caucasian vs African American or Hispanic), payer (private insurance vs Medicaid), cancer stage (advanced vs early), annual surgeon volume (low/medium [2–9 surgeries per year] or high [>10 surgeries per year] vs very low [1 surgery per year]), and surgeon specialty (gynecologic oncologists vs obstetrician gynecologists or general surgeons). Among nonteaching hospitals, medium-volume hospitals (10–19 ovarian cancer surgeries per year) and high-volume hospitals (≥20 surgeries per year) had significantly higher comprehensive surgery rates than low-volume facilities (1–9 surgeries per year). Volume did not influence comprehensive surgery rates in teaching hospitals.


Many women with ovarian cancer, especially those in poor, elderly, or minority groups, are not receiving recommended comprehensive surgery. Efforts should be made to ensure that all women with ovarian cancer, especially those in vulnerable populations, have the opportunity to receive care from centers or surgeons with higher comprehensive surgery rates. Cancer 2007. © 2007 American Cancer Society.

Ovarian cancer is the leading cause of death from gynecologic malignancies in the United States. Each year, approximately 24,000 women are diagnosed with ovarian cancer, and 14,300 die of the disease.1 The poor prognosis is related largely to the fact that approximately 70% of epithelial tumors, the most common type of ovarian cancer, are diagnosed at an advanced stage. It has not been demonstrated that screening for ovarian cancer to detect early-stage disease reduces its morbidity or mortality.2–5 Optimizing treatment strategies, such as complete surgical removal of the tumor, currently is the most effective way to improve ovarian cancer outcomes. Five-year survival rate for patients with advanced disease who receive optimal surgical cytoreduction (no residual disease) is from 30% to 40%, and the rate is only from 0% to 15% with suboptimal cytoreduction.1

Investigators have demonstrated that characteristics of the ovarian cancer surgeon and treating hospital also may affect prognosis. Women with ovarian cancer, especially at advanced stages, who are treated by gynecologic oncologists have significantly longer survival than women who are not.6–11 Junor et al.8 demonstrated that patients with stage III ovarian cancer who received surgical treatment by gynecologic oncologists had a 25% reduction in 5-year mortality risk. This benefit is associated with the higher rate at which gynecologic oncologists perform complete surgical cytoreduction. Receipt of care from high-volume and teaching hospitals also is associated with the provision of recommended ovarian cancer treatment, which, in turn, is associated with improved survival.11–13

Despite these findings, little is known about the health systems within which ovarian cancer patients in the United States receive their care. Two recently published studies from single states reported that the minority of ovarian cancer patients received care from gynecologic oncologists or from high-volume hospitals or surgeons.6, 14 This was confirmed in 2 additional studies that evaluated approximately 3000 women from the Surveillance, Epidemiology, and End Results (SEER)-Medicare database.15, 16 More broadly representative, population-based studies examining the health systems within which ovarian cancer patients receive care are important to ensure that all patients with ovarian cancer receive the most effective treatment.

The objective of the current study was to evaluate the surgical treatment received by patients with ovarian cancer in 9 states and to identify the patient, physician, ecologic, and hospital-based factors associated with receipt of comprehensive surgical treatment. In 2002, the American College of Obstetricians and Gynecologists issued a “Committee Opinion” that outlined the importance of referral to a gynecologic oncologist when ovarian cancer is suspected.17 However, currently, to our knowledge, there is no published information on the percentage of patients with ovarian cancer in the United States who receive surgical care from physicians who perform high volumes of ovarian cancer surgery, the majority of whom are gynecologic oncologists. Our study also addresses this important issue.



A comprehensive analytic study database linked patient-, surgeon-, hospital-, and county-level data from various sources. Patient-level data were obtained from the Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project's (HCUP's) state inpatient databases (SIDs), which contain discharge abstracts for all inpatient stays from 39 states across the United States.18 The SID is comprised of individual data files with discharge information abstracted by hospitals in the participating states. These discharge abstracts include patient demographics, county or zip code of residence, ≥10 diagnostic and 6 to 10 procedural International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM) codes, and payer for these hospitalizations. Selected states' databases include encrypted physician identifiers, including specialty for the primary surgeon, and a hospital identifier that facilitates linkage to the American Hospital Association Annual Survey of Hospitals Database.19 This database includes hospital average daily census and ownership as well as name, address, and zip code, which were used to link status as an obstetrics-gynecology teaching facility from the Association of Professors of Gynecology and Obstetrics' Directory of Residency Programs. Hospital, and patient zip codes were linked to Rural Urban Commuting Area (RUCA) codes, which classify zip codes into 1 of 30 categories based on population, location in an urban setting, and commuting patterns.20 We collapsed RUCAs into 3 categories—urban, large rural, and small/isolated small rural.21 We linked county-level data from the Area Resource File, which includes information on the number of generalist and subspecialist obstetrician-gynecologists in patient care.22 Finally, United States Census data on ecologic characteristics that represented socioeconomic status, education, and social class were linked to patient residence zip and Federal Information Processing Standard (FIPS) county codes.23

Study Population

The study included all 11,608 hospital admissions of women aged ≥21 years who were identified in the HCUP database as having received primary surgical treatment for ovarian cancer in 9 states over 3 years. We included the most recent 3 years of hospital discharge data available for these states (1999–2001 for Colorado, South Carolina, Wisconsin; 2000–2002 for Florida, Iowa, Maine, New Jersey, New York, Washington). Primary surgical treatment for ovarian cancer was defined as a primary ovarian cancer diagnosis (ICD-9-CM principal diagnosis code 183.0) and an ovarian resection (ICD-9 procedure codes 65.3–65.6) during their hospitalization. We excluded the admissions of 639 women who had a simultaneous, noncutaneous, second primary malignancy and the admissions of 537 women who lived outside the 9-state area, resulting in a final study population of 10,432 women.

Comprehensiveness of Ovarian Cancer Surgery

Diagnostic and procedural ICD-9-CM codes were used to assess comprehensiveness of surgery based on the National Institutes of Health (NIH) Ovarian Cancer Consensus Guidelines.3 Women who had comprehensive surgery underwent 1) both a lymph node dissection and omentectomy/cytoreduction, or 2) had a diagnosis of a secondary malignancy of a specified organ/site, such as the intestine or peritoneum, and underwent an omentectomy/cytoreduction (lymph node dissection was not considered essential for comprehensiveness if there was extensive disease outside the pelvis). All others were classified as having noncomprehensive surgery, partial staging, or indeterminate according to Table 1. Although we were not able to evaluate whether optimal cytoreduction occurred, we were able to determine whether the procedures recommended in the NIH Consensus Statement were performed. This measure represents the best proxy for comprehensive surgery using these data.

Table 1. Algorithm for Categorizing Ovarian Cancer Surgical Management
Surgical procedure beyond oophorectomyDiagnosis
Primary ovarian cancer onlyPrimary ovarian cancer with secondary malignancy of
Lymph node dissectionOmentectomy or cytoreductionResection of destruction of an organLymph nodesOther organsUnspecified
  1. NC indicates not comprehensive; PS, partial staging; I, indeterminate; C, comprehensive.


Explanatory Variables

Patient characteristics

Among the sociodemographic characteristics, age was available in all 9 study states; race (Asian, black, white, other) was available in 7 states, and Hispanic ethnicity was available in 6 states. Primary payer variables (from all 9 states) identified insurance status and served as a proxy for socioeconomic status.

The 30-indicator comorbidity measure published by Elixhauser et al.24 was used to identify admissions with preexisting comorbidities unrelated to the primary diagnosis of ovarian cancer. Acuity of the hospital admission (elective, urgent, emergent) measured case severity.

Cancer characteristics

Early-stage disease was defined as a primary ovarian cancer diagnosis in HCUP (ICD-9-CM code 183.0) with or without secondary cancer of the female genital tract (ICD-9-CM code 198.82). Advanced-stage disease was defined as primary ovarian cancer with secondary cancer of abdominal or iliac lymph nodes (ICD-9-CM codes 196.2 and 196.6), nongenital abdominal structures (ICD-9-CM codes 197.4–197.89), or pleura (ICD-9-CM code 197.2).

Ecologic characteristics

We linked age-specific median household income and percentage of women aged ≥25 years with a high school education within each admission's zip code or county as proxies for socioeconomic status, education, and social class. We calculated the county-based ratio of obstetrician-gynecologists to the population of women aged ≥20 years as a measure of gynecologic resource availability. The state served as a measure of regional variation, and the residence location type served as a measure of rurality.

Surgeon characteristics

An encrypted physician identifier that was available in 5 state databases allowed calculation of year-specific volume of ovarian cancer resections for each surgeon. Surgeon specialty was available in Maine and South Carolina; subspecialty to the gynecologic oncologist level was reported in South Carolina only.

Hospital characteristics

Hospital identifiers in all state databases except South Carolina allowed calculation of the year-specific volume of ovarian cancer resections for each hospital. Hospital location was identified as urban, large rural, or small/isolated small rural. Teaching hospitals were those that provided obstetrics-gynecology residency training.

Statistical Analysis

In bivariate analyses, chi-square tests were used to test for differences in surgical comprehensiveness among the 10,432 admissions with different patient, ecologic, hospital, and surgeon characteristics. Logistic regression analysis using SAS software, version 8.2, identified the characteristics that were associated independently with comprehensive surgical care. Our primary logistic regression model used data from the 5 states that had a full set of variables (n = 6854 admissions). Regardless of their significance, we included the variables age, race, stage, comorbidity, zip code-based median household income, and state in our models. Additional variables were included if they were significant predictors of our study outcome (P ≤ .05 with Wald chi-square test) or if they improved the fit of the stepwise regression model at a significance level ≤.05. We applied generalized estimating equation using identity-link function methods to account for clustering of admissions by surgeon and hospital and observed essentially no difference from the original regression results. Thus, we present the original regression findings. Because our outcome—receipt of comprehensive surgical care—is common in the study population, the adjusted odds ratio derived from the logistic regression does not approximate the relative risk (RR). Using predicted values from logistic regressions on 1000 samples with replacement, we computed the mean RR and 95% confidence interval (95% CI) around the RR by using standard methods.25 Alternate models that included data from >5 states but excluded independent variables that were not available in those states' databases confirmed that our findings were robust.


This study included 10,432 admissions of women who were hospitalized for both a primary diagnosis of ovarian cancer and an oophorectomy. The mean age was 60.1; 36.3% of patients had Medicare insurance (Table 2). Among the women for whom race was known, 83.6% were Caucasian. Advanced disease was coded in 61.1% of admissions, less than the 70% typically reported. Because advanced staging requires appropriate surgery and accurate coding, this number probably is artificially low. Twenty-four percent of admissions were urgent or emergent. The most common comorbidity was hypertension (26.1%) followed by fluid and electrolyte disorders (12.3%).

Table 2. Comprehensive Surgery Rates by Sociodemographic, Ecologic, and Clinical Characteristics of Ovarian Cancer Admissions*
CharacteristicNo.PercentageRate of comprehensive surgery
  • SD indicates standard deviation.

  • *

    Missing values: Race was missing for all admissions in Maine (186) and Washington (955) and for some admissions in Colorado (199), Florida (87), Iowa (151), New Jersey (43), New York (310), and Wisconsin (22); residence location was missing for all admissions in South Carolina (469) and Maine (186) and for some admissions in Colorado (1), New Jersey (4), and New York (3). Also missing were values for primary payer (1), median household income (19), and acuity of admission (15).

  • Significant at P < .001.

  • Significant at P < .01.

  • §

    Median household income is at the ZIP code level for all states except Maine and South Carolina, which have county level data only.

  • Significant at P < .05.

  • Only those comorbid conditions identified among ≥5% of the population are listed. Elixhauser et al., 199824 defined 30 comorbid measures.

Sociodemographic characteristics
 Age (mean± SD, 60.1± 14.4), y
  Caucasian (non-Hispanic)708483.668
  African American5436.459.7
  Asian or Pacific Islander1331.660.2
 Primary payer
  Private insurance548752.668.7
Ecologic characteristics
 Residence location
  Large rural4564.767.1
  Small/isolated small rural9259.565.5
  New Jersey12551262
  New York293628.169.2
  South Carolina4694.564.4
 Year of procedure
 Median household income in ZIP code or county§
 Obstetrician-gynecologists/100 K population of women ≥20 years in county of residence
Clinical characteristics
 Cancer stage
 Acuity of admission
  Deficiency anemia9519.170.8
  Chronic pulmonary disease8167.865
  Diabetes, uncomplicated7026.766.1
  Fluid and electrolyte disorders128712.369.1

Of the 9963 admissions with hospital descriptors available, 42.3% of patients received their care in teaching hospitals (Table 3). One-third of patients underwent surgery in low-volume hospitals. It is noteworthy that 25.2% of patients underwent surgery by very-low-volume surgeons who performed only 1 ovarian cancer surgery annually, and 48.3% underwent surgery by physicians who performed <10 such surgeries per year. Gynecologic oncologists treated 69.6% of women in South Carolina.

Table 3. Distribution of Ovarian Cancer Admissions and Comprehensive Surgery Rates by Treating Hospital and Surgeon Characteristics*
CharacteristicNo. of hospitalsAdmissions
No.%Rate of comprehensive surgery
  • *

    Missing values: Hospital identifiers (and, thus, all hospital variables) were missing for all admissions in South Carolina (469); average daily census was missing for 1 additional admission in Florida; surgeon identifiers (and, thus, surgeon volume) were missing for all admissions in Maine (186), South Carolina (469), and Wisconsin (858) and for some admissions in Iowa (117), New Jersey (155), New York (69), and Washington (1).

  • Significant at P < .001.

  • Maine and South Carolina were the only states with surgeon specialty data; Maine does not include gynecologic oncology as a subspecialty; Maine had 3 admissions and South Carolina had 38 admissions with unspecified surgeon specialty. Because these states do not provide a surgeon identifier, the number of surgeons by specialty could be computed.

Teaching hospital
Annual ovarian cancer surgery volume
 Low (1–9/y)585332033.357.4
 Medium (10–19/y)51180118.169.5
 High (≥20/y)45484248.673.7
Teaching-volume combination
 Low volume, nonteaching534282328.356.4
 Low volume, teaching50496563.5
 Medium volume, nonteaching32112311.367.9
 Medium volume, teaching196786.872.1
 High volume, nonteaching16180518.174.5
 High volume, teaching29303730.573.2
 Large rural713513.556.1
 Small/isolated small rural1122712.746.5
 Government, nonfederal85131113.268.7
 For profit857147.268.6
Average daily census
 No. of surgeons
Annual surgeon ovarian cancer surgery volume
 Very low (1/y)1829216525.255.2
 Low/medium (2–9/y)400194422.765.1
 High (≥10/y)84446852.175.2
 Maine admissions
  No.%Rate of comprehensive surgery
Total 18310064.5
Surgeon specialty
 Gynecologic oncologist 
 General obstetrician-gynecologist 15986.968.6
 General surgeon 2413.137.5
 South carolina admissions
  No.%Rate of comprehensive surgery
Total 43110064
Surgeon specialty
 Gynecologic oncologist 30069.675.7
 General obstetrician-gynecologist 11827.437.3
 General surgeon 13338.5

Overall, 66.9% of admissions underwent comprehensive ovarian surgery. Crude rates of surgical comprehensiveness differed significantly by several patient, ecologic, hospital, and surgeon characteristics (Tables 2, 3). The oldest and youngest women were the least likely to receive comprehensive surgical care among different age groups. Caucasian women were most likely among racial/ethnic groups to receive comprehensive surgery. Colorado had the highest rate of comprehensive surgery, and Iowa had the lowest. Admissions of self-pay and Medicaid patients had the lowest rates among the different payer groups. Women in hospitals with obstetrics-gynecology teaching programs were more likely than women in nonteaching hospitals to receive comprehensive surgery. The higher the ovarian cancer surgery volume and the average daily census, the higher the rates of comprehensive care. Admissions of patients in urban hospitals had the highest rates of comprehensiveness among those in hospitals in different locations. Like in the hospitals, the higher the surgeon volume, the higher the rate of comprehensive surgical care. In South Carolina, women whose surgeons were gynecologic oncologists had the highest comprehensive surgery rates among women with surgeons of different specialties.

Logistic regression using data on the 6854 admissions from the 5 states with all variables of interest identified factors associated with comprehensiveness independent of confounding factors (Fig. 1). Table 4 shows full the regression results. Patient characteristics that were associated independently with lower rates of comprehensive surgery were age >71 years compared with ages 21 years to 50 year, African Americans and Hispanics compared with Caucasians, and having congestive heart failure or weight loss. Race remained significant even after controlling for median household income. Advanced-stage disease compared with early-stage disease and having hypertension and hypothyroidism were associated with higher rates of comprehensive surgery. The type of admission was not associated significantly with comprehensive surgery. Receiving care from a hospital in a small/isolated small rural area was associated with significantly lower rates of comprehensive surgical care compared with receiving treatment in urban areas. Admissions of patients who underwent surgery in nonteaching hospitals with medium and high ovarian cancer surgery volumes had significantly higher comprehensiveness rates than patients who underwent surgery low-volume hospitals. Among teaching hospitals, ovarian cancer surgery volume did not influence comprehensiveness rates. Surgeon volume also was associated with the provision of comprehensive surgical care. Admissions of patients cared for by high-volume surgeons (≥10 ovarian cancer surgeries per year) and by low/medium-volume surgeons (2–9 ovarian cancer surgeries per year) had higher surgical comprehensiveness rates than admissions of patients of very-low-volume surgeons.

Figure 1.

The adjusted relative risk of comprehensive surgery with 95% confidence intervals by demographic, ecologic, and clinical characteristics of ovarian cancer admissions. The logistic model used data from 5 states with information on hospital and surgeon volume and patient race: Colorado (n = 641 admissions), Florida (n = 2667 admissions), Iowa (n = 465 admissions), New Jersey (n = 1255 admissions), and New York (n = 2936 admissions). Missing data for model variables included race (n = 790 admissions), median household income (n = 17 admissions), hospital location (n = 1 admissions), and surgeon volume (n = 341 admissions). The final model contained 6854 admissions after accounting for the missing variables within the 5 states. An asterisk indicates that, to examine the volume-comprehensiveness correlation within teaching and nonteaching hospitals separately, different reference groups were used. For the nonteaching hospital analysis, low-volume, nonteaching hospitals were used as the reference group. For the teaching hospital analysis, low-volume teaching hospitals were used as the reference group.

Table 4. Final Logistic Regression Model of Predictors of Comprehensive Surgical Care for Ovarian Cancer (n = 6854 Admissions)*
CharacteristicOR (95% CI)
  • OR indicates odds ratio; 95% CI, 95% confidence interval; Ref, reference group.

  • *

    The logistic model used data from the 5 states with information on hospital and surgeon volume and patient race: Colorado (n = 641), Florida (n = 2667), Iowa (n = 465), New Jersey (n = 1255), and New York (n = 2936). Missing data for model variables included race (790), median household income (17), hospital location (1), and surgeon volume (341). The final model contained 6854 admissions after accounting for the missing variables within the 5 states.

  • The list of comorbidities was derived from the list published by Elixhauser et al., 1998.24

Age, y
 21–501.00 (Ref)
 51–601.07 (0.92–1.26)
 61–700.88 (0.74–1.05)
 71–800.79 (0.64–0.97)
 ≥810.54 (0.41–0.72)
 Caucasian (non-Hispanic)1.00 (Ref)
 African American0.66 (0.52–0.83)
 Hispanic0.76 (0.60–0.95)
 Asian or Pacific Islander0.66 (0.44–0.99)
 Other0.85 (0.64–1.14)
Cancer stage
 Early1.00 (Ref)
 Advanced4.78 (4.26–5.37)
 Alcohol abuse1.22 (0.44–3.38)
 Deficiency anemia0.99 (0.82–1.20)
 Rheumatoid arthritis/collagen vascular disease0.84 (0.50–1.42)
 Blood loss anemia0.91 (0.54–1.53)
 Congestive heart failure0.65 (0.47–0.89)
 Chronic pulmonary disease0.87 (0.71–1.07)
 Coagulopathy1.10 (0.71–1.69)
 Depression1.00 (0.72–1.40)
 Diabetes, uncomplicated0.94 (0.75–1.18)
 Diabetes, complicated0.92 (0.42–2.04)
 Drug abuse1.13 (0.27–4.68)
 Hypertension1.24 (1.08–1.42)
 Hypothyroidism1.26 (1.02–1.56)
 Liver disease0.82 (0.47–1.45)
 Fluid and electrolyte disorders0.95 (0.80–1.13)
 Other neurologic disorders1.20 (0.75–1.92)
 Obesity0.89 (0.62–1.26)
 Paralysis1.80 (0.62–5.21)
 Peripheral vascular disorders1.33 (0.66–2.68)
 Psychoses0.85 (0.50–1.46)
 Pulmonary circulation disorders1.12 (0.50–2.54)
 Renal failure0.68 (0.30–1.55)
 Valvular disease1.16 (0.87–1.54)
 Weight loss0.68 (0.49–0.97)
Median household income
 ≤$25,0001.00 (Ref)
 $25,001–$35,0000.99 (0.81–1.20)
 $35,001–$45,0000.96 (0.77–1.19)
 ≥$45,0011.03 (0.83–1.28)
 Colorado1.00 (Ref)
 Florida0.65 (0.50–0.84)
 Iowa0.72 (0.48–1.09)
 New Jersey0.63 (0.48–0.84)
 New York0.80 (0.61–1.04)
Location of hospital
 Urban1.00 (Ref)
 Large rural0.97 (0.67–1.39)
 Small/isolated small rural0.60 (0.41–0.88)
Obstetrician-gynecologists per 100 K population in county of residence
 01.00 (Ref)
 1–100.75 (0.46–1.23)
 11–500.97 (0.67–1.40)
 ≥510.79 (0.53–1.18)
Teaching status and hospital ovarian cancer volume
 Low volume, nonteaching1.00 (Ref)
 Low volume, teaching1.30 (1.00–1.68)
 Medium volume, nonteaching1.66 (1.30–2.10)
 Medium volume, teaching1.37 (1.11–1.67)
 High volume, nonteaching1.31 (1.10–1.56)
 High volume, teaching1.62 (1.32–1.99)
Surgeon ovarian cancer volume
 Very low (1/y)1.00 (Ref)
 Low/medium (2–9/y)1.35 (1.15–1.58)
 High (≥10/y)1.57 (1.34–1.85)

We explored several alternative regression models. The first examined the relation between surgeon and hospital volume independent of hospital teaching status (Table 5). Admissions of patients of high-volume surgeons had similar surgical comprehensiveness rates regardless of hospital volume. For admissions of patients of medium- and low-volume surgeons, however, increasing hospital volume was associated with higher surgical comprehensiveness rates.

Table 5. Unadjusted Comprehensiveness Rates and Adjusted Relative Risk of Comprehensiveness by Hospital and Surgeon Volume*
Hospital volumeSurgeon volume
Very low (1/year)Low (2–9/Year)High (%≥10/year)
Unadjusted rateAdjusted RR (95% CI)Unadjusted rateAdjusted RR (95% CI)Unadjusted rateAdjusted RR (95% CI)
  • RR indicates relative risk; 95% CI, 95% confidence interval.

  • *

    Calculated among 8577 admissions with both hospital and surgeon volume variable data available.

  • Missing values: Surgeon identifiers (and, thus, surgeon volume) were missing for all admissions in Maine (186), South Carolina (469), and Wisconsin (858) and for some admissions in Iowa (117), New Jersey (155), New York (69), and Washington (1).

  • Missing values: Hospital identifiers (and, thus, hospital volume) were missing for all admissions in South Carolina (469).

Low (1–9/y)
Medium (10–19/y)62.71.26 (1.13–1.39)66.81.18 (1.05–1.31)73.90.99 (0.84–1.14)
High (≥20/y)61.71.43 (1.06–1.92)71.51.49 (1.12–1.98)75.71.01 (0.76–1.34)

We used separate regression models to evaluate the factors associated with surgical comprehensiveness stratified by early stage and late-stage disease. Age, race, and surgeon volume remained independent predictors for both groups. Hospital volume in nonteaching hospitals was an independent predictor among admissions of early-stage patients, but not advanced-stage patients. Among teaching hospitals, volume did not have an impact on the likelihood of comprehensive surgical care for admissions of either early stage or late-stage patients. Because nearly all admissions of patients aged ≥65 years are insured by Medicare, insurance status was correlated highly with age. Thus, we only included the payer variable in a regression analysis that was limited to admissions of patients ages <65 years. This model indicated that Medicaid admissions were significantly less likely than privately insured admissions to receive comprehensive surgical care (RR, 0.89; 95% CI, 0.79–0.98).

In a model that included only South Carolina and Maine, where specialty data were available, obstetrician-gynecologists were significantly more likely to perform comprehensive surgical care than general surgeons (RR, 1.72; 95% CI, 1.16–2.15). In South Carolina only, where we could determine specialty to the gynecologic oncologist level, both general surgeons (RR, 0.34; 95% CI, 0.11–0.77) and obstetrician gynecologists (RR, 0.54; 95% CI, 0.37–0.73) were significantly less likely to perform comprehensive surgery than gynecologic oncologists. However, we were not able to control for hospital or surgeon volumes in these models, because they are not reported in South Carolina.


The Institute of Medicine's recent publication, Ensuring Quality Cancer Care, drew national attention to the quality of cancer care and concluded that many Americans with cancer are not receiving ideal care.26 For ovarian cancer, which often is diagnosed at advanced stages, the key to high-quality care is appropriate surgical treatment.27 However, little is known about the health systems within which patients with ovarian cancer in the United States receive their treatment and whether they are providing the recommended, comprehensive surgical treatment. This study has taken a first step toward filling this information gap by identifying health-system factors associated with the receipt of recommended surgical treatment for ovarian cancer.

Although the hospital discharge data provided an opportunity to examine patterns of ovarian cancer surgical care in 9 states, we were limited by the information available in these databases. Several states do not report variables, such as race or hospital and surgeon identifiers. In addition, individual patients cannot be followed on different admissions to a hospital. Therefore, a patient may appear once in the database as undergoing noncomprehensive surgery and then may appear again as a different patient undergoing a more comprehensive procedure. However, because we required the inclusion of women who underwent oophorectomy, this should represent the minority of women. Although clinical detail is lost when the complexity of a patient's health and treatment is reduced to a series of diagnosis and procedure codes used to support billing, discharge data can facilitate a practical and cost-effective study of health care utilization, especially in large populations. Other database limitations include 1) the lack of linked mortality data beyond the hospitalization to examine outcome, 2) limited information on patients' comorbidities, and 3) lack of pathologic information. In addition, ICD-9-CM codes do not distinguish between omentectomy and cytoreduction. Thus, for this study, we examined the association between health systems and the most basic staging and treatment procedures, but we could not determine whether or not cytoreduction specifically took place. This definition of comprehensive surgical care makes our findings of lower comprehensiveness rates among some health systems even more compelling, because it probably likely over-represents the comprehensiveness of ovarian cancer care. Despite these limitations, our findings are consistent with a growing body of literature indicating the importance of hospital and surgeon volume as predictors of the receipt of optimal care.

Over the past 20 years, numerous studies28–31 have shown that high-volume surgeons and hospitals have significantly lower 30-day and 60-day mortality rates of performing complex surgical procedures for conditions like carcinoma of the pancreas, esophagus, prostate, bladder, lung, and colon. Investigating the postoperative mortality-volume relation for cancer surgery in California, Dudley et al.32 concluded that 602 postoperative deaths could have been avoided if all patients, with 11 high-risk conditions, had been treated at high-volume centers. For less complex procedures, the volume-postoperative mortality relation is less likely to be significant31–33 Research results on volume and longer term outcomes as well as cure rates are much more limited. In a review of urologic cancer research, provider volume was related inversely to cancer progression and surgical complications.34 In studies of breast cancer, both hospital and surgeon volume as well as surgical oncology specialty were associated with higher 5-year survival rates35, 36 In a population-based study from Canada that evaluated the influence of volume on ovarian cancer outcomes, hospital and surgeon volume were associated inversely with need for reoperation within 3 months of initial surgery.37 Surgical specialty influenced 30-day mortality, reoperation, and overall survival. In 2 recently published studies evaluating a population of women covered by Medicare who were identified in the SEER registry from 1992 to 1999, hospital volume and surgeon volume were not strong determinants of ovarian cancer surgical outcomes, but surgeon specialty influenced surgical process, short-term mortality, and overall survival.15, 16

Others have found that teaching hospitals are more likely to provide appropriate care for women with ovarian cancer. Harlan et al. reported that women who were treated in a hospital with an approved residency were significantly more likely to receive guideline therapy and appropriate surgery.12 Tingulstad et al. reported that patients with ovarian cancer who were treated at a teaching hospital in Norway had significantly lower risk of death compared with patients who were treated at other centers,11 independent of the chemotherapy received, suggesting that the higher rate of optimal cytoreduction may be responsible for the differences in survival. In England, investigators observed that inappropriate surgery was more likely to be performed in nonteaching hospitals.38 However, those studies did not control for hospital volume.

The current study extended the work of those investigators by examining the interactions between teaching status, hospital volume, and surgical volume, and we observed a complex relation between them. Hospital teaching status was not an independent predictor of comprehensive surgery. The relation between hospital volume and comprehensive surgery rates varied according to the hospital's teaching status. Hospital volume influenced comprehensive surgical care rates in nonteaching hospitals, but not in teaching hospitals. Further analysis stratifying admissions by cancer stage demonstrated that, in nonteaching hospitals, volume was predictive of comprehensive surgical care in women with early-stage disease, but not in women with advanced-stage disease.

Our results also indicated that a provider's surgical volume was an important predictor of the provision of comprehensive ovarian cancer surgery. Surgeon volume, unlike hospital volume, was important in the care of both early stage and advanced-stage ovarian cancer, suggesting that surgeon volume is a more important predictor of comprehensive surgery. Some investigators who examined the surgeon-hospital volume relation reported that individual surgeon volume was more important than hospital volume in predicting operative mortality and postoperative morbidity,30 but most investigators observed that both variables were independently significant.39 However, there is substantial variation between surgeons. One study of high-volume surgeons who performed radical prostatectomy demonstrated huge variations in postoperative complications and long-term morbidity, indicating the importance of individual surgeon technique.40

In addition to volume, we observed that specific patient and ecologic factors were associated with receiving comprehensive surgical care. Elderly women (aged >70 years), African Americans, and Hispanics all were less likely to receive comprehensive surgery than young women (ages 21–50 years) and Caucasians. Those women who were treated in a small/isolated small rural setting were less likely to receive comprehensive surgery. Many others have demonstrated disparities in the availability and utilization of cancer care across different ages, racial and ethnic groups, income groups, and geographic locations.41–45 Few studies have focused on ovarian cancer; however, Harlan et al. used 1991 and 1996 SEER data supplemented with medical records to evaluate adherence to the NIH guidelines for ovarian cancer treatment and reported findings consistent with ours.12, 46 Those women who were older, African American, Hispanic, and lacked private insurance all had significantly lower rates of guideline therapy and significantly higher rates of inappropriate surgery. Our current findings suggest that not much progress has been made in closing these gaps in the past decade. We also observed that there were significant differences in comprehensiveness rates based on state (Table 4), but this did not correlate with the state density of gynecologic oncologists. Using 2000 Census data and the Society of Gynecologic Oncologists database, the approximate number of gynecologic oncologists per 1000 residents in each state was 1/172 for New York, 1/245 for Florida, 1/250 for New Jersey, 1/333 for South Carolina, 1/347 for Washington, 1/383 for Colorado, 1/385 for Wisconsin, 1/433 for Maine, and 1/483 for Iowa.

Another objective of the current study was to assess the health systems within which ovarian cancer patients received care. For ovarian cancer surgery, 33% of women were treated in low-volume hospitals, and 25% were treated by surgeons who only performed 1 ovarian cancer surgery annually. However, 52% were treated by high-volume surgeons who performed ≥10 ovarian cancer surgeries per year. Bristow et al. evaluated volume-based patterns of care in Maryland and observed that 50% of women were treated in low-volume hospitals and that only 34% were treated by high-volume surgeons (≥10 surgeries per year); 91% of the surgeons who treated ovarian cancer performed, on average, only 1 ovarian cancer surgery per year.14 In a population-based study from Utah, only 39% of ovarian cancer patients were treated by a gynecologic oncologist.6 In both of those studies, age and rural residence were associated negatively with receiving care by gynecologic oncologists or high-volume surgeons. Schrag et al. observed that, in a population-based study of Medicare patients, 49% of surgeries were performed by surgeons who only performed such operations occasionally.15 Because optimal surgery with cytoreduction is associated with improved survival, efforts should be made to ensure that all women with ovarian cancer, especially those who are vulnerable because of age, race, or socioeconomic status, are referred to centers or surgeons from whom they are more likely to get optimal surgery. Our study findings suggest that the referral of women with suspected ovarian cancer to expert centers for primary surgery would be an effective strategy to improve overall outcomes for women with ovarian cancer.


Supported in part by the Centers for Disease Control and Prevention (CDC) and the National Cancer Institute through the Cancer Prevention and Control Research Network, a network within the CDC's Prevention Research Centers Program (1-U48-DP-000050).