Abstract

BACKGROUND: The presence and type of health insurance may be an important determinant of cancer stage at diagnosis. To determine whether previously observed racial differences in stage of cancer at diagnosis may be explained partly by differences in insurance coverage, we studied all patients with incident cases of melanoma or colorectal, breast, or prostate cancer in Florida in 1994 for whom the stage at diagnosis and insurance status were known. METHODS: The effects of insurance and race on the odds of a late stage (regional or distant) diagnosis were examined by adjusting for an individual's age, sex, marital status, education, income, and comorbidity. All P values are two-sided. RESULTS: Data from 28 237 patients were analyzed. Persons who were uninsured were more likely diagnosed at a late stage (colorectal cancer odds ratio [OR] = 1.67, P = .004; melanoma OR = 2.59, P = .004; breast cancer OR = 1.43, P = .001; prostate cancer OR = 1.47, P = .02) than were persons with commercial indemnity insurance. Patients insured by Medicaid were more likely diagnosed at a late stage of breast cancer (OR = 1.87, P <.001) and melanoma (OR = 4.69, P <.001). Non-Hispanic African-American patients were more likely diagnosed with late stage breast and prostate cancers than were non-Hispanic whites. Hispanic patients were more likely to be diagnosed with late stage breast cancer but less likely to be diagnosed with late stage prostate cancer. CONCLUSIONS: Persons lacking health insurance and persons insured by Medicaid are more likely diagnosed with late stage cancer at diverse sites, and efforts to improve access to cancer-screening services are warranted for these groups. Racial differences in stage at diagnosis are not explained by insurance coverage or socioeconomic status.

Stage at diagnosis is one of the most important prognostic factors for most cancers ( 1 ). For many cancers, early stage disease can be effectively treated with good chance for cure, whereas late stage disease is generally incurable. Therefore, understanding the determinants of cancer stage at diagnosis is an important objective to improve cancer outcomes.

Stage at diagnosis is associated with a number of factors, including race and ethnicity ( 14 , 15 ). Several studies have suggested that the presence and type of health insurance may be important determinants. A more favorable stage at diagnosis has been associated with health maintenance organization (HMO) coverage ( 16 , 17 ), whereas Medicaid recipients and the uninsured have been more likely to be diagnosed with later stage cancer ( 18 , 19 ). The presence and type of health insurance have been consistently predictive of access to care and the provision of screening services ( 20 - 27 ).

Several questions remain unanswered, however, concerning the determinants of stage at diagnosis. Most studies reporting racial differences in stage at diagnosis have had limited information on health insurance ( 4-6,9,12,13 ) or were conducted in selected urban populations ( 2,3,8 ). Therefore, the extent to which racial differences in stage at diagnosis may be attributed to differences in insurance coverage and vice versa is not clear, and it is not clear whether racial differences exist on a statewide basis.

In addition, little is known about how specific types of health insurance coverage predict the stage of cancer at diagnosis. For example, most studies examining the effects of HMOs have been confined to the Medicare population. Whether HMOs have similar effects in non-Medicare populations and whether other insurance types (such as preferred provider organizations) affect stage is not known.

This study was conducted to examine the effects of health insurance coverage, race, and ethnicity on cancer stage at diagnosis by using population-based data for the state of Florida. We hypothesized that patients having HMO insurance would have a more favorable stage at diagnosis and that patients lacking health insurance or insured by Medicaid would be diagnosed at later stages relative to patients with commercial indemnity insurance.

P atients and M ethods

Data Sources

We studied data from 1994 (the most recent year for which all relevant data were available) for all Florida patients with incident cases of four types of cancer for which screening is associated with detection of early stage disease: colorectal, breast (female only), prostate, and melanoma (n = 34 616) ( 28 - 36 ). Cervical cancer was not included because of different reporting requirements for this site. ( In situ cervical cancers are not reportable.) Incident cases were identified from the Florida Cancer Data System (FCDS), Florida's population-based statewide cancer registry. The FCDS was created in 1978 and has been collecting data on cancer incidence since 1981. By state law, all primary malignant tumors that are in situ or invasive are reportable to FCDS except for in situ carcinoma of the cervix. The FCDS has well-established methods to ensure complete case finding, including cooperative arrangements with other state tumor registries, linkage with other databases, and ad hoc audits of reporting facilities.

For the inclusion of information that is not routinely available from the FCDS (insurance payer, comorbidity, and socioeconomic status), patient records were linked with state discharge abstracts and the 1990 U.S. Census. The state of Florida Agency for Health Care Administration (AHCA) maintains both inpatient and outpatient discharge abstracts for admissions to all non-Federal acute care hospitals, all licensed ambulatory surgical centers, all free-standing radiation therapy centers, and all diagnostic imaging centers. Data abstracted include patient-identifying information and demographics (e.g., Social Security number, date of birth, sex, race, and ethnicity), discharge diagnoses (up to 10), and insurance payer. The 1990 U.S. Census was used to obtain aggregate measures of socioeconomic status either by census tract or, if unavailable, by ZIP code.

This study was approved by the University of South Florida Institutional Review Board.

Matching Process

Records of FCDS patients were linked with discharge abstracts by means of a probabilistic match using Social Security number, sex, race, and date of birth. Patients who successfully matched on all variables were considered to be valid matches. Patients were also considered valid matches if the sole discrepancy was a Social Security number or a date of birth that differed by only one digit (suggesting data entry errors). By use of this method, 28 123 eligible patients (81.2%) were matched, a rate comparable to that achieved in a similar study ( 18 ).

Compared with successful matches, FCDS patients who failed to match with AHCA records were more likely to be non-Hispanic African-Americans or of “other” race ( P = .001), to be male ( P = .001), to be of older age ( P <.001), to have melanoma or prostate cancer ( P = .001), and to be diagnosed at the in situ stage ( P = .001). There was no difference in community-level income ( P = .08) or education ( P = .82) between matched and unmatched patients. Patients whom we failed to match were likely diagnosed in an outpatient setting and did not require hospital admission for diagnosis or follow-up treatment.

Information on patient's insurance payer was obtained from AHCA discharge data for patients who matched. For patients who did not match with AHCA discharge abstracts, we used payer information from the FCDS if it was available. Insurance payer is defined similarly by both FCDS and AHCA as the patient's primary method of payment or type of insurance coverage for services provided by the reporting facility. Information on the patient's insurance was available for the following percentage of patients: colon cancer patients, 89.9%; melanoma patients, 80.9%; breast cancer patients, 89.9%; and prostate cancer patients, 84.5%.

FCDS patients were next matched with 1990 U.S. Census data to obtain census-derived measures of socioeconomic status. The U.S. census provides data on median household income and median education level (according to race and ethnicity) for each census tract and ZIP code. Each individual was assigned the median income/education level of their racial category for either the census tract (87% of patients) or ZIP code (13% of patients) of their residence. Associations between socioeconomic status and health outcomes have been similar when socioeconomic status is measured at either the individual or the community level ( 37 , 38 ). Some researchers ( 39 , 40 ) have also argued that the socioeconomic level of a neighborhood is itself an important determinant of health.

There are important limitations, however, in inferring the socioeconomic status of individuals from grouped data. First, census-derived measures of socioeconomic status will not accurately reflect the individual education or income of some patients. In addition, in a rapidly growing state such as Florida, it is possible that some census tracts or ZIP codes may have had substantial changes in median education and income levels between 1990 and 1994. If so, 1990 census estimates may not be accurate measures of socioeconomic status for those patients diagnosed in 1994.

Data Analysis

Stage at diagnosis, the main outcome variable, was defined by using the Surveillance, Epidemiology, and End Results (SEER) 1 Site-Specific Summary Staging Guide ( 41 ). Stage at diagnosis is based on a combination of pathologic, surgical, and clinical assessments available within 2 months of diagnosis. For these analyses, stage at diagnosis was reclassified as early stage ( in situ or local) or late stage (regional or distant).

Stage at diagnosis was available for 32 077 FCDS patients (92.7% of all patients as follows: colon cancer, 93.5%; melanoma, 93.6%; breast cancer, 95.5%; and prostate cancer, 88.9%). Compared with staged cases of cancer, unstaged cases of cancer were found in older individuals ( P <.001) who had lower education ( P <.001) and income ( P <.001); who were less likely to be married ( P = .001); and who were more likely to be insured by Medicare ( P = .001), to be nonwhite ( P = .001), to be male ( P = .001), and to have prostate cancer ( P = .001).

Two variables were examined as possible predictors of stage at diagnosis: insurance payer and race. Categories of insurance payer are mutually exclusive and include the following: Medicare; Medicare HMO; Medicaid; Medicaid HMO; commercial indemnity; commercial preferred provider organization; commercial HMO; Civilian Health and Medical Program, Uniformed Service (CHAMPUS); U.S. Department of Veterans Affairs; workers' compensation; other state and local government programs; or uninsured (includes self-pay and charity). Because of the small number of Medicaid HMO patients (n = 71), this group was combined with Medicaid fee-for-service patients into a single Medicaid category. Likewise, the following payer types were combined under the category “other” because of the small number of patients: CHAMPUS, U.S. Department of Veterans Affairs, workers' compensation, and other state/local government programs. Patients were classified into the following four categories of race and ethnicity: white (non-Hispanic), African-American (non-Hispanic), Hispanic, or other.

Variables available directly from the FCDS that were examined as possible confounders included sex, age, and marital status (never married, married, divorced, separated, or widowed). Race-specific median household income is reported by the U.S. census on a 9-point scale. Because there were very few patients at the upper and lower extremes, this scale was collapsed to five levels for multivariate analysis: (less than $15 000, $15 000-$24 999, $25 000-$34 999, $35 000-$49 999, and $50 000 or more). Median educational attainment is classified by the U.S. census on a 7-point scale. Because there were very few patients at the upper and lower extremes, this scale was collapsed to four levels for multivariate analysis (less than high school graduate, high school graduate, some college, and college graduate).

Comorbidity was determined as described by Charlson et al. ( 42 ) and Deyo et al. ( 43 ). Comorbid conditions were identified by use of inpatient and outpatient hospital discharge abstracts for the calendar year 1994. We used the International Classification of Diseases, 9 th revision, Clinical Modification ( 44 ), mapping of comorbid conditions as described by Deyo et al. ( 43 ). Cancer-related conditions were excluded. In calculating a morbidity index (theoretic range, 0-23), we used the original weights described by Charlson et al. ( 42 ). We defined three categories of comorbidity (0, 1, and ⩾2) based on the patient's index score.

All analyses were conducted separately by site. The proportion of patients whose disease was diagnosed at a late stage (regional or distant) was first compared across payer and racial categories with the χ 2 test. The multivariate relationship between a late stage at diagnosis and race and payer was then examined by use of multiple logistic regression. Indicator variables were created for the four categories of race and the eight payer categories with white non-Hispanic and commercial indemnity (fee for service) insurance as the referent categories. Based on previous research, the following variables were included in all logistic models to control for potential confounding: age, sex (for colorectal cancer and melanoma), marital status (four indicator variables), educational level, income level, and comorbidity ( 2,13,14 ). Race and insurance payer indicator variables were likewise included in all logistic models so that the odds ratios (ORs) reported for race are adjusted for insurance payer and vice versa.

To avoid forcing linear relationships between ordinal and interval variables and outcomes, we first assessed these variables with Mantel-Haenszel's χ 2 test for trend ( 45 ). For those variables that did not demonstrate a monotonic relationship with the outcome variable, indicator variables were created (eight indicator variables for age corresponding to 10-year age intervals, three indicator variables for level of education, and four indicator variables for level of income) ( 46 ). The statistical significance of individual indicator variables was assessed with the χ 2 likelihood ratio test. Adjusted ORs and 95% confidence intervals (CIs) are reported. All reported P values are two-sided. Statistical significance was set at an alpha level of .05.

Initial logistic regression models examined the main effects of payer type and race. Interactions between race and insurance type were then examined with the Breslow-Day χ 2 test for homogeneity of ORs. For ORs that were not homogeneous, interaction terms were tested for statistical significance. To determine how findings might change under different assumptions, we repeated logistic models and restricted them to those patients with invasive cancers only and to those age groups for which screening is recommended by American Cancer Society guidelines (age ⩾50 years for colorectal cancer, age ⩾40 years for breast cancer, age ⩾50 years for prostate cancer, and age ⩾20 years for melanoma) ( 47 ). In addition, because age is a key determinant for some insurance payers (e.g., Medicare), logistic regression models were repeated stratified by age (age <65 years versus age ⩾ 65 years).

R esults

Our final study population consisted of the 28 237 Florida residents diagnosed with colorectal cancer, breast cancer, melanoma, or prostate cancer in 1994, for whom information was available on both stage and insurance payer (Table 1 ). Most patients were older than 65 years of age, and Medicare was the most common type of insurance. Table 2 presents the proportion of patients who were diagnosed at a late stage of cancer for each category of insurance payer and race. The insurance payer was statistically significantly associated with stage at diagnosis for each of the four cancers examined. Patients insured by Medicaid and patients who were uninsured were at greater risk for late stage disease. Non-Hispanic African-Americans were at a greater risk of a late stage diagnosis for breast and prostate cancers.

Table 3 reports adjusted ORs, comparing the odds of a late stage diagnosis for insurance payer and race categories relative to referent categories. Final logistic models are described separately below for each site.

Colorectal cancer. Two payer categories were found to have statistically significantly greater odds of late stage disease than the referent category of commercial indemnity insurance: Medicare HMO and uninsured. Patients insured by a Medicare HMO also had statistically significantly greater odds of late stage disease when compared with patients insured by fee-for-service Medicare as an alternate referent category (adjusted OR = 1.59; 95% CI = 1.28-1.97; P <.001). There was a trend for a more favorable stage at diagnosis for persons of “other” race compared with the referent of non-Hispanic whites. The effects of Medicare HMO ( P = .49) and uninsured payer status ( P = .062) were both homogeneous across categories of race.

Melanoma. Two payer categories were associated with later stage at diagnosis: Medicaid and the uninsured. The number of cases of late stage melanoma was quite small for several insurance categories, including Medicaid (n = 13), Medicare HMO (n = 6), and patients having other insurance (n = 11). Very few cases of melanoma were diagnosed in nonwhite patients, and there were no statistically significant associations between race and stage at diagnosis. ORs for Medicaid ( P = .14) and uninsured payer status ( P = .18) were homogeneous across categories of race.

Breast cancer. Patients insured by Medicaid and those who were uninsured were more likely to be diagnosed with late stage disease. Patients who were non-Hispanic African-American or Hispanic were more likely to be diagnosed at a late stage independent of their insurance payer. There was a statistically nonsignificant trend for patients of “ other” race to have a more favorable stage at diagnosis. The effects of non-Hispanic African-American race ( P = .92) and Hispanic ethnicity ( P = .37) were homogeneous across payer categories. Likewise, ORs for Medicaid ( P = .59) and uninsured payer status ( P = .69) were homogeneous across categories of race.

Prostate cancer. Patients who were uninsured had greater odds of a late stage diagnosis than those with commercial indemnity coverage. There was a trend toward a less favorable stage at diagnosis for patients insured by Medicaid. Non-Hispanic African-American patients had a less favorable stage at diagnosis, whereas patients who were Hispanic had a more favorable stage than non-Hispanic whites, the referent group. The effects of Hispanic ethnicity on stage at diagnosis were not homogeneous across payer categories ( P = .03), and the effects of being uninsured were not homogeneous across categories of ethnicity ( P = .004). We therefore tested the statistical significance of an interaction between Hispanic ethnicity and being uninsured in logistic models. This interaction was statistically significant (OR = 3.12; 95% CI = 1.47-6.40; P = .003). Hispanic ethnicity was associated with the diagnosis at an earlier stage of disease if patients were insured (OR = 0.74; 95% CI = 0.61-0.91; P = .005) but was associated with diagnosis at a later stage of disease if uninsured (OR = 2.32; 95% CI = 1.13-4.76; P = .02). Being uninsured had no significant effect on non-Hispanic whites (OR = 1.20; 95% CI = 0.84-1.73; P = .32) but greatly increased the odds of late stage disease for Hispanics (OR = 3.75; 95% CI = 1.90-7.36; P <.001).

For each of the four cancer types examined, results were similar to those described above when logistic models were repeated with data restricted to those patients with invasive cancers only. In addition, results did not vary when restricted to age groups for whom screening is recommended or when analyses were stratified by age (<65 years old versus ⩾65 years old; data not presented).

D iscussion

For all four types of cancer examined, the presence and type of health insurance were statistically significantly associated with stage at diagnosis. Our finding of increased odds of a late stage diagnosis among patients who were uninsured or who were insured by Medicaid is in agreement with other studies ( 2,3,8,18,19 ). The previous demonstrations of a relationship between a lack of health insurance and poor cancer outcomes have primarily been limited to breast cancer. Our study extends this finding to other cancers that can be detected at an early stage by screening (notably, melanoma, colorectal cancer, and prostate cancer).

A plausible explanation for a later stage at diagnosis among the uninsured is inadequate access to cancer-screening services. Persons lacking health insurance are less likely to receive cancer screening and other preventive services ( 21,22,27,48 ) and may have delayed diagnosis of symptomatic cancers because of inadequate access to health care ( 7 , 10 ). Until means are found to provide health insurance to all persons, it appears inevitable that the uninsured will suffer worse outcomes for cancer and other diseases ( 25 , 49 ).

The reason for a later stage at diagnosis among patients insured by Medicaid compared with those who have indemnity insurance is less clear and warrants further study. Presumably, patients insured by Medicaid have access to cancer screening and diagnostic services. We found, however, that the odds of a late stage cancer diagnosis were often similar between Medicaid patients and patients who were uninsured. Medicaid may not have provided the continuous insurance coverage that is conducive to comprehensive preventive care. One study ( 50 ), for example, found that almost two thirds of new Medicaid recipients lost coverage within 12 months. Whether Medicaid HMOs have outcomes that are different from fee-for-service Medicaid care is unknown because there were too few patients from Medicaid HMOs in this study to allow meaningful comparisons.

We found no difference in stage at diagnosis for patients insured by commercial HMOs compared with fee-for-service health plans. Among previous investigations that examined stage at diagnosis ( 51 - 56 ), HMO plans were usually no different than fee-for-service plans. Studies showing a more favorable stage at diagnosis for HMO patients ( 16,17,57 ) have primarily been confined to Medicare populations receiving care in well-established staff model HMOs. Some investigators ( 58 - 60 ) have questioned whether Independent Practice Association (IPA) forms of HMO provide the same quality of care as staff-model HMOs.

The explanation for our finding of greater risk of late stage colorectal cancer among Medicare HMO patients is uncertain and merits further exploration. Some early Medicare HMO plans in Florida were plagued with problems ranging from poor quality of care to outright fraud. IMC Gold Plus, the nation's largest prepaid Medicare plan in 1987, was one such example ( 61 , 62 ). Similar problems were uncovered among Florida Medicaid HMOs in 1994 ( 63 , 64 ). We are not aware of evidence, however, that Medicare HMOs available in Florida during 1994 were suspected of providing poor quality care; moreover, patients in Medicare HMOs were not at higher risk of late stage diagnosis for the other cancers examined.

There may be financial disincentives, however, for primary care physicians to perform colorectal examinations such as sigmoidoscopy under capitated health plans ( 65 ). Capitated primary care physicians receive a set payment for all health care services that they provide and are unlikely to receive additional reimbursement for the time it takes them to educate and encourage patients to undergo colorectal screening or for the time it takes them to perform sigmoidoscopy in their office. Other screening tests, such as mammography, prostate-specific antigen assays, and skin examinations, are less likely to have differing financial implications for primary care physicians treating fee-for-service versus HMO patients.

As a general model, insurance-related differences in stage at diagnosis should be greater for cancers that are more amenable to early detection and where early detection health services are more dependent on the presence of health insurance (expensive tests for example). The effects of being uninsured or of having Medicaid were fairly consistent across the four cancers examined. A lack of statistical significance for data from colorectal and prostate cancer patients insured by Medicaid may partly be the result of smaller sample sizes. The particularly striking insurance effects among patients with melanoma suggest that melanoma is especially amenable to early detection or that patients without adequate health insurance may have markedly limited access to dermatologists. Although skin examinations are relatively inexpensive, skin biopsies and other follow-up care may not be affordable to patients lacking health insurance.

We observed racial differences in stage at diagnosis for both breast and prostate cancers independent of patient's insurance coverage, education, and income. Studies ( 2,3,8-10,12,13 ) have consistently identified racial differences in stage at diagnosis and survival among patients with breast cancer. Although African-American women have been found to receive less breast cancer screening than white women ( 21,66,67 ), stage differences in breast cancer by race may only be partly explained by inadequate screening or other problems in health care delivery ( 4,7,68,69 ).

Less is known about racial differences in stage at diagnosis for other cancers. Patients who are African-American may have a later stage at diagnosis for colon cancer ( 1 , 70 ) and prostate cancer ( 71 ). Results for Hispanic patients have been inconsistent; some results suggest earlier stage at diagnosis for breast, colon, or prostate cancer ( 72 ), and other results show a greater likelihood of late stage breast cancer ( 10 ). Our study found that Hispanic patients are diagnosed at a later stage for breast cancer but at an earlier stage for prostate cancer. We also found that effects of Hispanic ethnicity were modified by the presence of health insurance, which may partly explain the variation in results from previous studies.

This study has a number of potential limitations. First, we relied on administrative data that are limited in detail (no information on type of HMO or TNM [ tumor-node-metastasis] staging system) and that could not be independently verified. Because cancers are increasingly diagnosed and treated in outpatient settings, case finding for cancers such as melanoma may be incomplete. We did not have information on the duration or continuity of insurance coverage before diagnosis. If cancer-related symptoms led persons to change their health insurance coverage, this change could have affected our results ( 73 ). Finally, our study was restricted to patients in Florida with incident cases of cancer, so findings might not be generalizable to other parts of the country.

In conclusion, race and insurance status are important determinants of cancer stage at diagnosis. Patients who are uninsured or who are insured by Medicaid are more likely to be diagnosed with late stage cancers, as were persons of African-American race for breast and prostate cancers. Further research is needed to confirm and understand potential differences between HMO and fee-for-service health care systems in the detection of early stage colorectal cancer.

Table 1.

Characteristics of men and women (n = 28 237) diagnosed with selected cancers in Florida, 1994

Characteristic *
 
Colorectal cancer (n = 8090)
 
Melanoma (n = 1531)
 
Breast cancer (n = 9888)
 
Prostate cancer (n = 8728)
 
Mean age, y 71.6 62.3 63.9 69.4 
Median household income $28 861 $31 441 $29 582 $29 488 
Sex, No. (%) 
 Male 4129 (51.1) 917 (59.9)  8728 (100) 
 Female 3958 (48.9) 614 (40.1) 9888 (100) 
Race, No. (%) 
 White, non-Hispanic 6928 (85.6) 1439 (94.0) 8306 (84.0) 7011 (80.3) 
 African-American, non-Hispanic 464 (5.7) 14 (0.9) 677 (6.8) 762 (8.7) 
 Hispanic 636 (7.9) 53 (3.5) 757 (7.7) 877 (10.0) 
 Other 62 (0.8) 25 (1.6) 148 (1.5) 78 (0.9) 
Education, No. (%) 
 High school graduate or less 3801 (47.3) 535 (35.2) 4275 (43.5) 4004 (46.2) 
 More than high school education 4238 (52.8) 989 (65.0) 5557 (56.5) 4657 (53.8) 
Marital status, No. (%) 
 Never 564 (7.1) 153 (10.4) 864 (8.9) 525 (6.2) 
 Current 4891 (61.7) 1052 (71.2) 5570 (57.6) 6925 (82.0) 
 Divorced/separated 499 (6.3) 95 (6.4) 930 (9.6) 404 (4.8) 
 Widowed 1983 (25.0) 177 (12.0) 2305 (23.8) 588 (7.0) 
Insurance payer, No. (%) 
 Medicare 5284 (65.3) 669 (43.7) 4511 (45.6) 5431 (62.2) 
 Medicare HMO 452 (5.6) 64 (4.2) 401 (4.1) 306 (3.5) 
 Medicaid 119 (1.5) 28 (1.8) 249 (2.5) 96 (1.1) 
 Commercial indemnity 709 (8.8) 274 (17.9) 1653 (16.7) 957 (11.0) 
 Commercial HMO 662 (8.2) 180 (11.8) 1081 (10.9) 864 (9.9) 
 Commercial preferred provider organization 484 (6.0) 186 (12.1) 1161 (11.7) 596 (6.8) 
 Uninsured 234 (2.9) 72 (4.7) 472 (4.8) 245 (2.8) 
 Other 146 (1.8) 58 (3.8) 360 (3.6) 233 (2.7) 
Stage, No. (%) 
In situ 347 (4.3) 176 (11.5) 953 (9.6) 31 (0.4) 
 Local 2606 (32.2) 1131 (73.9) 5924 (59.9) 6945 (79.6) 
 Regional 3759 (46.5) 119 (7.8) 2475 (25.0) 1246 (14.3) 
 Distant 1378 (17.0) 105 (6.9) 536 (5.4) 506 (5.8) 
Comorbid condition(s), No. (%) 
 0 5455 (67.4) 1401 (91.5) 8513 (86.1) 7323 (83.9) 
 1 1929 (23.8) 106 (6.9) 1120 (11.3) 1099 (12.6) 
 ⩾2 706 (8.7) 24 (1.6) 255 (2.6) 306 (3.5) 
Characteristic *
 
Colorectal cancer (n = 8090)
 
Melanoma (n = 1531)
 
Breast cancer (n = 9888)
 
Prostate cancer (n = 8728)
 
Mean age, y 71.6 62.3 63.9 69.4 
Median household income $28 861 $31 441 $29 582 $29 488 
Sex, No. (%) 
 Male 4129 (51.1) 917 (59.9)  8728 (100) 
 Female 3958 (48.9) 614 (40.1) 9888 (100) 
Race, No. (%) 
 White, non-Hispanic 6928 (85.6) 1439 (94.0) 8306 (84.0) 7011 (80.3) 
 African-American, non-Hispanic 464 (5.7) 14 (0.9) 677 (6.8) 762 (8.7) 
 Hispanic 636 (7.9) 53 (3.5) 757 (7.7) 877 (10.0) 
 Other 62 (0.8) 25 (1.6) 148 (1.5) 78 (0.9) 
Education, No. (%) 
 High school graduate or less 3801 (47.3) 535 (35.2) 4275 (43.5) 4004 (46.2) 
 More than high school education 4238 (52.8) 989 (65.0) 5557 (56.5) 4657 (53.8) 
Marital status, No. (%) 
 Never 564 (7.1) 153 (10.4) 864 (8.9) 525 (6.2) 
 Current 4891 (61.7) 1052 (71.2) 5570 (57.6) 6925 (82.0) 
 Divorced/separated 499 (6.3) 95 (6.4) 930 (9.6) 404 (4.8) 
 Widowed 1983 (25.0) 177 (12.0) 2305 (23.8) 588 (7.0) 
Insurance payer, No. (%) 
 Medicare 5284 (65.3) 669 (43.7) 4511 (45.6) 5431 (62.2) 
 Medicare HMO 452 (5.6) 64 (4.2) 401 (4.1) 306 (3.5) 
 Medicaid 119 (1.5) 28 (1.8) 249 (2.5) 96 (1.1) 
 Commercial indemnity 709 (8.8) 274 (17.9) 1653 (16.7) 957 (11.0) 
 Commercial HMO 662 (8.2) 180 (11.8) 1081 (10.9) 864 (9.9) 
 Commercial preferred provider organization 484 (6.0) 186 (12.1) 1161 (11.7) 596 (6.8) 
 Uninsured 234 (2.9) 72 (4.7) 472 (4.8) 245 (2.8) 
 Other 146 (1.8) 58 (3.8) 360 (3.6) 233 (2.7) 
Stage, No. (%) 
In situ 347 (4.3) 176 (11.5) 953 (9.6) 31 (0.4) 
 Local 2606 (32.2) 1131 (73.9) 5924 (59.9) 6945 (79.6) 
 Regional 3759 (46.5) 119 (7.8) 2475 (25.0) 1246 (14.3) 
 Distant 1378 (17.0) 105 (6.9) 536 (5.4) 506 (5.8) 
Comorbid condition(s), No. (%) 
 0 5455 (67.4) 1401 (91.5) 8513 (86.1) 7323 (83.9) 
 1 1929 (23.8) 106 (6.9) 1120 (11.3) 1099 (12.6) 
 ⩾2 706 (8.7) 24 (1.6) 255 (2.6) 306 (3.5) 
*

Numbers for individual categories may not sum to total sample size because of missing data and rounding. HMO = health maintenance organization.

Table 2.

Percentage of patients (n = 28 237) diagnosed at a late stage of disease by site

Predictor variable *
 
No. of patients diagnosed with cancer at late stage (%)
 
Colon cancer (n = 8090)
 
Melanoma (n = 1531)
 
Breast cancer (n = 9888)
 
Prostate cancer (n = 8728)
 
Insurance payer 
 Medicare 3248 (61.5) 86 (12.9) 1187 (26.3) 1062 (19.6) 
 Medicare HMO 322 (71.2) 6 (9.4) 113 (28.2) 65 (21.2) 
 Medicaid 88 (73.9) 13 (46.4) 130 (52.2) 27 (28.1) 
 Commercial indemnity 468 (66.0) 37 (13.5) 525 (31.8) 193 (20.2) 
 Commercial HMO 423 (63.9) 28 (15.6) 365 (33.8) 165 (19.1) 
 Commercial preferred provider organization 316 (65.3) 20 (10.8) 375 (32.3) 128 (21.5) 
 Uninsured 179 (76.5) 23 (31.9) 201 (42.6) 67 (27.4) 
 Other 93 (63.7) 11 (19.0) 115 (31.9) 45 (19.3) 
  Total 5137 (63.5) 224 (14.6) 3011 (30.5) 1752 (20.1) 
  χ 2 (two-sided P value)   χ 2 = 46.4   χ 2 = 46.6   χ 2 = 135.1   χ 2 = 14.5  
   for payer differences  (.001) (.001) (.001) (.044) 
Race, No. (%) 
 White 4365 (63.0) 209 (14.5) 2390 (28.8) 1383 (19.7) 
 African-American 322 (69.4) 4 (28.60) 313 (46.2) 199 (26.1) 
 Hispanic 416 (65.4) 8 (15.1) 273 (36.1) 154 (17.6) 
 Other 34 (54.8) 3 (12.0) 35 (23.7) 16 (20.5) 
  Total 5137 (63.5) 224 (14.6) 3011 (30.5) 1752 (20.1) 
  χ 2 (two-sided P value)   χ 2 = 10.7   χ 2 = 2.3   χ 2 = 105.1   χ 2 = 21.3  
   for race differences  (.013) (.505) (.001) (.001) 
Predictor variable *
 
No. of patients diagnosed with cancer at late stage (%)
 
Colon cancer (n = 8090)
 
Melanoma (n = 1531)
 
Breast cancer (n = 9888)
 
Prostate cancer (n = 8728)
 
Insurance payer 
 Medicare 3248 (61.5) 86 (12.9) 1187 (26.3) 1062 (19.6) 
 Medicare HMO 322 (71.2) 6 (9.4) 113 (28.2) 65 (21.2) 
 Medicaid 88 (73.9) 13 (46.4) 130 (52.2) 27 (28.1) 
 Commercial indemnity 468 (66.0) 37 (13.5) 525 (31.8) 193 (20.2) 
 Commercial HMO 423 (63.9) 28 (15.6) 365 (33.8) 165 (19.1) 
 Commercial preferred provider organization 316 (65.3) 20 (10.8) 375 (32.3) 128 (21.5) 
 Uninsured 179 (76.5) 23 (31.9) 201 (42.6) 67 (27.4) 
 Other 93 (63.7) 11 (19.0) 115 (31.9) 45 (19.3) 
  Total 5137 (63.5) 224 (14.6) 3011 (30.5) 1752 (20.1) 
  χ 2 (two-sided P value)   χ 2 = 46.4   χ 2 = 46.6   χ 2 = 135.1   χ 2 = 14.5  
   for payer differences  (.001) (.001) (.001) (.044) 
Race, No. (%) 
 White 4365 (63.0) 209 (14.5) 2390 (28.8) 1383 (19.7) 
 African-American 322 (69.4) 4 (28.60) 313 (46.2) 199 (26.1) 
 Hispanic 416 (65.4) 8 (15.1) 273 (36.1) 154 (17.6) 
 Other 34 (54.8) 3 (12.0) 35 (23.7) 16 (20.5) 
  Total 5137 (63.5) 224 (14.6) 3011 (30.5) 1752 (20.1) 
  χ 2 (two-sided P value)   χ 2 = 10.7   χ 2 = 2.3   χ 2 = 105.1   χ 2 = 21.3  
   for race differences  (.013) (.505) (.001) (.001) 
*

HMO = health maintenance organization.

Late stage is defined as regional or distant.

Two-sided P value for χ 2 test.

Table 3.

Summary of logistic regressions: adjusted odds ratios (ORs) (95% confidence intervals [CIs]) for late stage diagnosis (n = 28 053)

Variable *
 
OR (95% CI)
 
Colon cancer (n = 8035)
 
Melanoma (n = 1524)
 
Breast cancer (n = 9835)
 
Prostate cancer (n = 8659)
 
Insurance payer 
 Commercial indemnity 1.00 (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) 
 Medicare 0.93 (0.77-1.13) 1.04 (0.61-1.79) 0.98 (0.83-1.14) 1.03 (0.85-1.24) 
 Medicare HMO 1.48 (1.13-1.94)  0.72 (0.27-1.91) 1.05 (0.81-1.36) 1.14 (0.83-1.59) 
 Medicaid 1.44 (0.92-2.25) 4.69§ (1.90-11.56)  1.87§ (1.41-2.45)  1.56 (0.96-2.54) 
 Commercial HMO 0.96 (0.76-1.21) 1.20 (0.69-2.08) 1.06 (0.90-1.25) 0.96 (0.76-1.21) 
 Commercial preferred provider organization 0.94 (0.74-1.21) 0.74 (0.41-1.33) 0.98 (0.83-1.15) 1.06 (0.82-1.36) 
 Other 0.91 (0.63-1.33) 1.38 (0.64-2.94) 1.02 (0.79-1.31) 0.95 (0.66-1.37) 
 Uninsured 1.67 (1.18-2.36)  2.59 (1.37-4.90)  1.43 (1.15-1.77)  1.47# (1.06-2.04)  
Race 
 Non-Hispanic white 1.00 (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) 
 Non-Hispanic African-American 1.13 (0.89-1.44) 1.93 (0.54-6.86) 1.65§ (1.39-1.97)  1.27# (1.05-1.54)  
 Hispanic 1.07 (0.89-1.29) 0.70 (0.30-1.65) 1.26** (1.07-1.49)  0.80# (0.66-0.97)  
 Other 0.60 (0.36-1.02) 0.85 (0.24-3.08) 0.71 (0.48-1.05) 1.06 (0.61-1.85) 
Variable *
 
OR (95% CI)
 
Colon cancer (n = 8035)
 
Melanoma (n = 1524)
 
Breast cancer (n = 9835)
 
Prostate cancer (n = 8659)
 
Insurance payer 
 Commercial indemnity 1.00 (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) 
 Medicare 0.93 (0.77-1.13) 1.04 (0.61-1.79) 0.98 (0.83-1.14) 1.03 (0.85-1.24) 
 Medicare HMO 1.48 (1.13-1.94)  0.72 (0.27-1.91) 1.05 (0.81-1.36) 1.14 (0.83-1.59) 
 Medicaid 1.44 (0.92-2.25) 4.69§ (1.90-11.56)  1.87§ (1.41-2.45)  1.56 (0.96-2.54) 
 Commercial HMO 0.96 (0.76-1.21) 1.20 (0.69-2.08) 1.06 (0.90-1.25) 0.96 (0.76-1.21) 
 Commercial preferred provider organization 0.94 (0.74-1.21) 0.74 (0.41-1.33) 0.98 (0.83-1.15) 1.06 (0.82-1.36) 
 Other 0.91 (0.63-1.33) 1.38 (0.64-2.94) 1.02 (0.79-1.31) 0.95 (0.66-1.37) 
 Uninsured 1.67 (1.18-2.36)  2.59 (1.37-4.90)  1.43 (1.15-1.77)  1.47# (1.06-2.04)  
Race 
 Non-Hispanic white 1.00 (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) 
 Non-Hispanic African-American 1.13 (0.89-1.44) 1.93 (0.54-6.86) 1.65§ (1.39-1.97)  1.27# (1.05-1.54)  
 Hispanic 1.07 (0.89-1.29) 0.70 (0.30-1.65) 1.26** (1.07-1.49)  0.80# (0.66-0.97)  
 Other 0.60 (0.36-1.02) 0.85 (0.24-3.08) 0.71 (0.48-1.05) 1.06 (0.61-1.85) 
*

HMO = health maintenance organization.

ORs for late stage diagnosis (regional/distant stage) are adjusted for age, education, income, sex (where appropriate), marital status, race, insurance payer, and comorbidity by use of multiple logistic regression. ORs for insurance payer are adjusted for race and vice versa. ORs with two-sided P value <.05 are in boldface type.

Two-sided P = .005.

§

Two-sided P <.001.

Two-sided P = .004.

Two-sided P = .001.

#

Two-sided P = .02.

**

Two-sided P = .006.

1

Editor's note: SEER is a set of geographically defined, population-based, central cancer registries in the United States, operated by local nonprofit organizations under contract to the National Cancer Institute (NCI). Registry data are submitted electronically without personal identifiers to the NCI on a biannual basis, and the NCI makes the data available to the public for scientific research.

R. G. Roetzheim and J. Z. Ayanian were supported through Generalist Physician Faculty Scholars Awards from the Robert Wood Johnson Foundation.

We thank Jill MacKinnon for her assistance in obtaining Florida Cancer Data System data and Dr. Alan Cantor for his review of this report.

References

(1)
Kosary CL, Ries LA, Miller BA, Hankey BF, Harras A, Edwards BK. SEER cancer statistics review, 1973-1992: tables and graphs. Bethesda (MD): National Cancer Institute; 1995.
(2)
Hunter CP, Redmond CK, Chen VW, Austin DF, Greenberg RS, Correa P, et al. Breast cancer: factors associated with stage at diagnosis in black and white women. Black/White Cancer Survival Study Group.
J Natl Cancer Inst
 
1993
;
85
:
1129
-37.
(3)
Eley JW, Hill HA, Chen VW, Austin DF, Wesley MN, Muss HB, et al. Racial differences in survival from breast cancer. Results of the National Cancer Institute Black/White Cancer Survival Study.
JAMA
 
1994
;
272
:
947
-54.
(4)
Jones BA, Kasl SV, Curnen MG, Owens PH, Dubrow R. Can mammography screening explain the race difference in stage at diagnosis of breast cancer?
Cancer
 
1995
;
75
:
2103
-13.
(5)
Gordon NH, Crowe JP, Brumberg DJ, Berger NA. Socioeconomic factors and race in breast cancer recurrence and survival.
Am J Epidemiol
 
1992
;
135
:
609
-18.
(6)
Mandelblatt J, Andrews H, Kao R, Wallace R, Kerner J. Impact of access and social context on breast cancer stage at diagnosis.
J Health Care Poor Underserved
 
1995
;
6
:
342
-51.
(7)
Coates RJ, Bransfield DD, Wesley M, Hankey B, Eley JW, Greenberg RS, et al. Differences between black and white women with breast cancer in time from symptom recognition to medical consultation. Black/White Cancer Survival Study Group.
J Natl Cancer Inst
 
1992
;
84
:
938
-50.
(8)
Perkins P, Cooksley CD, Cox JD. Breast cancer. Is ethnicity an independent prognostic factor for survival?
Cancer
 
1996
;
78
:
1241
-7.
(9)
Bassett MT, Krieger N. Social class and blackwhite differences in breast cancer survival.
Am J Public Health
 
1986
;
76
:
1400
-3.
(10)
Richardson JL, Langholz B, Bernstein L, Burciaga L, Danley K, Ross RK. Stage and delay in breast cancer diagnosis by race, socioeconomic status, age and year.
Br J Cancer
 
1992
;
65
:
922
-6.
(11)
Farley TA, Flannery JT. Late-stage diagnosis of breast cancer in women of lower socioeconomic status: public health implications.
Am J Public Health
 
1989
;
79
:
1508
-12.
(12)
Wells BL, Horm JW. Stage at diagnosis in breast cancer: race and socioeconomic factors.
Am J Public Health
 
1992
;
82
:
1383
-5.
(13)
Mandelblatt J, Andrews H, Kerner J, Zauber A, Burnett W. Determinants of late stage diagnosis of breast and cervical cancer: the impact of age, race, social class, and hospital type [published erratum appears in Am J Public Health 1991;81:980].
Am J Public Health
 
1991
;
81
:
646
-9.
(14)
Nayeri K, Pitaro G, Feldman JG. Marital status and stage at diagnosis in cancer.
N Y State J Med
 
1992
;
92
:
8
-11.
(15)
Moritz DJ, Satariano WA. Factors predicting stage of breast cancer at diagnosis in middle aged and elderly women: the role of living arrangements.
J Clin Epidemiol
 
1993
;
46
:
443
-54.
(16)
Riley GF, Potosky AL, Lubitz JD, Brown ML. Stage of cancer at diagnosis for Medicare HMO and fee-for-service enrollees.
Am J Public Health
 
1994
;
84
:
1598
-604.
(17)
Potosky AL, Merrill RM, Riley GF, Taplin SH, Barlow W, Fireman BH, et al. Breast cancer survival and treatment in health maintenance organization and fee-for-service settings.
J Natl Cancer Inst
 
1997
;
89
:
1683
-91.
(18)
Ayanian JZ, Kohler BA, Abe T, Epstein AM. The relation between health insurance coverage and clinical outcomes among women with breast cancer.
N Engl J Med
 
1993
;
329
:
326
-31.
(19)
Osteen RT, Winchester DP, Hussey DH, Clive RE, Friedman MA, Cady B, et al. Insurance coverage of patients with breast cancer in the 1991 Commission on Cancer Patient Care Evaluation Study.
Ann Surg Oncol
 
1994
;
1
:
462
-7.
(20)
Blustein J. Medicare coverage, supplemental insurance, and the use of mammography by older women.
N Engl J Med
 
1995
;
332
:
1138
-43.
(21)
Breast and cervical cancer screening among underserved women. Baseline survey results from six states. The National Cancer Institute Cancer Screening Consortium for Underserved Women.
Arch Fam Med
 
1995
;
4
:
617
-24.
(22)
Hayward RA, Shapiro MF, Freeman HE, Corey CR. Who gets screened for cervical and breast cancer? Results from a new national survey.
Arch Intern Med
 
1988
;
148
:
1177
-81.
(23)
Kirkman-Liff B, Kronenfeld JJ. Access to cancer screening services for women.
Am J Public Health
 
1992
;
82
:
733
-5.
(24)
Kaplan RM, Navarro AM, Castro FG, Elder JP, Mishra SI, Hubbell A, et al. Increased use of mammography among Hispanic women: baseline results from the NCI Cooperative Group on Cancer Prevention in Hispanic Communities.
Am J Prev Med
 
1996
;
12
:
467
-71.
(25)
Schroeder SA. The medically uninsured—will they always be with us?
N Engl J Med
 
1996
;
334
:
1130
-3.
(26)
Valdini A, Cargill LC. Access and barriers to mammography in New England community health centers.
J Fam Pract
 
1997
;
45
:
243
-9.
(27)
Woolhandler S, Himmelstein DV. Reverse targeting of preventive care due to lack of health insurance.
JAMA
 
1988
;
259
:
2872
-4.
(28)
Catalona WJ, Richie JP, Ahmann FR, Hudson MA, Scardino PT, Flanigan RC, et al. Comparison of digital rectal examination and serum prostate specific antigen in the early detection of prostate cancer: results of a multicenter clinical trial of 6630 men.
J Urol
 
1994
;
151
:
1283
-90.
(29)
Mettlin CJ, Murphy GP, McGinnis LS, Menck HR. The National Cancer Data Base report on prostate cancer. American College of Surgeons Commission on Cancer and the American Cancer Society.
Cancer
 
1995
;
76
:
1104
-12.
(30)
Mettlin C, Murphy GP, Babaian RJ, Chesley A, Kane RA, Littrup PJ, et al. The results of a five-year early prostate cancer detection intervention. Investigators of the American Cancer Society National Prostate Cancer Detection Project.
Cancer
 
1996
;
77
:
150
-9.
(31)
Morrison AS. Review of evidence on the early detection and treatment of breast cancer.
Cancer
 
1989
;
64(12 Suppl)
:
2651
-6.
(32)
Shapiro S, Venet W, Strax P, Venet L. Periodic screening for breast cancer—The Health Insurance Plan project and its sequelae, 1963-1986. Baltimore (MD): Johns Hopkins University Press; 1988.
(33)
Newcomb PA, Norlfleet RG, Storer BE, Surawicz TS, Marcus PM. Screening sigmoidoscopy and colorectal cancer mortality.
J Natl Cancer Inst
 
1992
;
84
:
1572
-5.
(34)
Selby JV, Friedman GD, Quesenberry CP Jr, Weiss NS. A case-control study of screening sigmoidoscopy and mortality from colorectal cancer.
N Engl J Med
 
1992
;
326
:
653
-7.
(35)
Berwick M, Begg CB, Fine JA, Roush GC, Barnhill RL. Screening for cutaneous melanoma by skin self-examination.
J Natl Cancer Inst
 
1996
;
88
:
17
-23.
(36)
Roush GC, Berwick M, Koh HK, MacKie R. Screening for melanoma. Philadelphia (PA): JB Lippincott; 1992.
(37)
Hofer TP, Wolfe RA, Tedeschi PJ, McMahon LF, Griffith JR. Use of community versus individual socioeconomic data predicting variation in hospital use.
Health Serv Res
 
1998
;
33
:
243
-59.
(38)
Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology.
Am J Public Health
 
1992
;
87
:
703
-10.
(39)
Krieger N, Fee E. Social class: the missing link in U.S. health data.
Int J Health Serv
 
1994
;
24
:
25
-44.
(40)
Diez-Roux AV. Bringing context back into epidemiology: variables and fallacies in multilevel analysis.
Am J Public Health
 
1998
;
88
:
216
-22.
(41)
Shambaugh EM, Weiss MA. Summary staging guide: cancer Surveillance, Epidemiology, and End Results reporting. Bethesda (MD): National Institutes of Health; 1977.
(42)
Charlson ME, Pompei P, Ales KL, Mackenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.
J Chronic Dis
 
1987
;
40
:
373
-83.
(43)
Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases.
J Clin Epidemiol
 
1992
;
45
:
613
-9.
(44)
International Classification of Diseases, 9 th revision, clinical modification. Washington (DC): Public Health Service, US Department of Health and Human Services; 1980.
(45)
Hosmer DW, Lemeshow S. Applied logistic regression. New York (NY): John Wiley & Sons; 1989.
(46)
Greenland S. Modeling and variable selection in epidemiologic analysis.
Am J Public Health
 
1989
;
79
:
340
-9.
(47)
Cancer facts and figures. Atlanta (GA): American Cancer Society; 1998.
(48)
Screening mammography: a missed clinical opportunity? Results of the NCI Breast Cancer Screening Consortium and National Health Interview Survey Studies.
JAMA
 
1990
;
264
:
54
-8.
(49)
Blendon RJ, Benson JM. Whatever happened to politicians' concerns about the nation's uninsured? [editorial].
Am J Public Health
 
1998
;
88
:
345
-6.
(50)
Carrasquillo O, Himmelstein DU, Woolhandler S, Bor DH. Can Medicaid managed care provide continuity of care to new Medicaid enrollees? An analysis of tenure on Medicaid.
Am J Public Health
 
1998
;
88
:
464
-6.
(51)
Kulkarni PR, Vernon SW, Jackson GL, Lairson D, Davis BR. Stage at diagnosis of breast cancer. Comparison in a fee-for-service and health maintenance organization practice.
Med Care
 
1989
;
27
:
608
-22.
(52)
Greenwald HP. HMO membership, copayment, and initiation of care for cancer: a study of working adults.
Am J Public Health
 
1987
;
77
:
461
-6.
(53)
Greenwald HP, Henke CJ. HMO membership, treatment, and mortality risk among prostatic cancer patients.
Am J Public Health
 
1992
;
82
:
1099
-104.
(54)
Francis AM, Polissar L, Lorenz AB. Care of patients with colorectal cancer. A comparison of a health maintenance organization and fee-for-service practices.
Med Care
 
1984
;
22
:
418
-29.
(55)
Retchin SM, Brown B. Management of colorectal cancer in Medicare health maintenance organizations.
J Gen Intern Med
 
1990
;
5
:
110
-4.
(56)
Vernon SW, Hughes JI, Heckel VM, Jackson GL. Quality of care for colorectal cancer in a fee-for-service and health maintenance organization practice [ published erratum appears in Cancer 1994;74:3249].
Cancer
 
1992
;
69
:
2418
-25.
(57)
Riley GF, Potosky AL, Klabunde CN, Warren JL, Ballard-Barbash R. Stage at diagnosis and treatment patterns among older women with breast cancer: an HMO and fee-for-service comparison.
JAMA
 
1999
;
281
:
720
-6.
(58)
Kuttner R. Must good HMOs go bad? First of two parts: the commercialization of prepaid group health care.
N Engl J Med
 
1998
;
338
:
1558
-63.
(59)
Kuttner R. Must good HMOs go bad? Second of two parts: the search for checks and balances.
N Engl J Med
 
1998
;
338
:
1635
-9.
(60)
Retchin SM. Heterogeneity of health maintenance organizations and quality of care.
J Natl Cancer Inst
 
1997
;
89
:
1654
-5.
(61)
Hollman L, Nohlgren S. Founder of IMC indicted. In: St. Petersburg Times. St. Petersburg (FL); April 23, 1987. p. 1A.
(62)
Gentry C, Nickens T. More problems for IMC: Medicare cancels contract. In: St. Petersburg Times. St. Petersburg (FL); May 2, 1987. p. 1A.
(63)
State vows reform of Medicaid HMOs. In: Sun-Sentinel. Ft. Lauderdale (FL); December 2, 1995.
(64)
Dismantle failing Medicaid HMOs, congressman says. In: Sun-Sentinel. Ft. Lauderdale (FL); December 22, 1994.
(65)
Leider PJ, Solberg R, Nesbitt T. Family physician perception of economic incentives for the provision of office procedures.
Fam Med
 
1997
;
29
:
318
-20.
(66)
McCarthy BD, Yood MU, MacWilliam CH, Lee MJ. Screening mammography use: the importance of a population perspective.
Am J Prev Med
 
1996
;
12
:
91
-5.
(67)
Burns RB, McCarthy EP, Freund KM, Marwill SL, Shwartz M, Ash A, et al. Black women receive less mammography even with similar use of primary care.
Ann Intern Med
 
1996
;
125
:
173
-82.
(68)
Chang SW, Kerlikowske K, Napoles-Springer A, Posner SF, Sickles EA, Perez-Stable EJ. Racial differences in timeliness of follow-up after abnormal screening mammography.
Cancer
 
1996
;
78
:
1395
-402.
(69)
Lannin DR, Mathews HF, Mitchell J, Swanson MS, Swanson FH, Edwards MS. Influence of socioeconomic and cultural factors on racial differences in late-stage presentation of breast cancer.
JAMA
 
1998
;
279
:
1801
-7.
(70)
Cooper GS, Yuan Z, Landefeld CS, Rimm AA. Surgery for colorectal cancer: race-related differences in rates and survival among Medicare beneficiaries.
Am J Public Health
 
1996
;
86
:
582
-6.
(71)
Robbins AS, Whittemore AS, Van Den Eeden SK. Race, prostate cancer survival, and membership in a large health maintenance organization.
J Natl Cancer Inst
 
1998
;
90
:
986
-90.
(72)
Goodwin JS, Samet JM, Hunt WC. Determinants of survival in older cancer patients.
J Natl Cancer Inst
 
1996
;
88
:
1031
-8.
(73)
Morgan RO, Virnig BA, DeVito CA, Persily NA. The Medicare-HMO revolving door—the healthy go in and the sick go out.
N Engl J Med
 
1997
;
337
:
169
-75.