Kirsner, Robert S. MD, PhD*†‡§; Ma, Fangchao MD, PhD*; Fleming, Lora MD, PhD†; Trapido, Edward ScD*; Duncan, Robert PhD†; Federman, Daniel G. MD¶; Wilkinson, James D. MD, MPH†
Two of the commonest types of health care delivery systems in the United States are fee-for-service and managed care systems (including the health maintenance organizations [HMO]). Medicare's HMO program has grown rapidly in recent years: As of 1999, more than 6.5 million persons were enrolled, accounting for 17% of the beneficiary population.1–2 An important issue is whether, and how, the health care delivery system in which patients participate may affect patient care. Managed care systems have been developed to contain health care costs, but their effect on quality of care has not been clearly established.3 It has been reported that some differences in patient outcomes are associated with the type of health care delivery system in which they are enrolled.4–8
Among women, breast and cervical cancer represent 2 of the most common cancer sites. Breast cancer is the most common nonskin cancer among women in the United States, second only to lung cancer as a cause of cancer-related death. In 2001, an estimated 190,000 new cases of breast cancer were diagnosed in American women, and more than 40,000 women died of the disease.9 The risk of developing breast cancer increases with age.10 Nearly 20,000 new cases of cervical cancer are diagnosed each year, and about 5,000 women die from this disease annually.11 The lifetime risk of dying from cervical cancer in the United States is 0.3%. Although the 5-year survival rate is about 90% for persons with localized cervical cancer, it is considerably lower (about 15%) for persons with advanced (stage IV) disease.
Routine screening exists for both breast and cervical cancer. There is evidence that mammography screening every 12–33 months significantly reduces mortality from breast cancer,12 particularly for women aged 50 years or older.13 For cervical cancer, regular screening with Pap tests are recommended for all women within 3 years of onset of sexual activity or age 21 (whichever comes first) and screening at least every 3 years and who have an intact cervix. Case-control studies have shown a strong negative association between cervical cancer screening and invasive disease, suggesting that cervical cancer screening is protective.14–15 The cumulative incidence of invasive cervical cancer was reduced by 64% when the interval between Pap tests was 10 years, by 84% at 5 year testing intervals, increasing to an incidence reduction of 94% when testing was performed annually.16
Using national health care databases, the current study evaluated differences in the stage of breast and cervical cancer at diagnosis between the 2 Medicare health care delivery systems (fee for service and HMO) during the period 1985–2001. We hypothesized that medicare-aged patients enrolled in HMOs would be diagnosed at an earlier stage; furthermore, because access (the ability to make use of) and actual use of care might be part of this reason, we reasoned that patients with a prior diagnosis of another cancer before their diagnosis of breast or cervical cancer might experience differences in stage at diagnosis. We therefore compared women both with and without a prior diagnosis of cancer to evaluate prior cancer diagnosis as a possible modifying factor to the association of type of health care delivery system and cancer stage at diagnosis.
MATERIALS AND METHODS
This study represents an analysis of the linkage of 2 national databases: the Medicare database from the Centers for Medicare and Medicaid Services (formerly the Health Care Financing Administration or HCFA) and the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results program database. The linked Surveillance, Epidemiology, and End Results Medicare data represent a large population-based source of information for cancer-related epidemiologic and health services research. The Surveillance, Epidemiology, and End Results Medicare data have been identified by the Institute of Medicine as one of the few population-based data resources available for analyses of the quality of cancer care.2
The Surveillance, Epidemiology, and End Results program is an epidemiologic surveillance system sponsored by the National Cancer Institute consisting of population-based tumor registries that routinely collect information on all newly diagnosed cancer (incident) cases that occur in persons residing in Surveillance, Epidemiology, and End Results areas.17 The information collected about each incident cancer diagnosis includes the patient's demographic characteristics, date of diagnosis, tumor data (eg, histology, stage, and grade), type of treatment recommended or provided within 4 months of diagnosis, follow-up of vital status, and cause of death, if applicable.
Although Surveillance, Epidemiology, and End Results data do not constitute a national probability sample, they are the primary source of national information on cancer incidence and survival.2 Reported information includes the month and year of diagnosis, stage at diagnosis, date of death, and county and census tract of residence.
Medicare enrollment files contain entitlement dates to Part A and Part B, ZIP code of residence, health care delivery type, and months in which the beneficiary was enrolled in a Medicare HMO.18 Medicare is the primary health insurer for 97% of the US population aged 65 years and older. Each Medicare file varies in the data elements included and the types of procedure and diagnostic codes used.
The linkage of the Surveillance, Epidemiology, and End Results and Medicare data is the result of the collaborative effort of the NCI, the Surveillance, Epidemiology, and End Results registries, and the Centers for Medicare and Medicaid Services. The linkage of the Surveillance, Epidemiology, and End Results data with the Medicare data entails an algorithm-based record match on social security number, name, sex, and date of birth that has been described previously.19 The linkage was initially completed in 1991 and has been updated twice, in 1995 and 1999. For each linked data set, among persons in the Surveillance, Epidemiology, and End Results data who were aged 65 years or older, 94% were matched to the Medicare enrollment database.2 Linkages are updated every 3 years. The most recent linkage, completed in 2002, has Surveillance, Epidemiology, and End Results cases through 1999 linked with Medicare claims data through 2001.
To ensure confidentiality, the Surveillance, Epidemiology, and End Results Medicare data records have been unidentified (ie, all personal identifiers have been removed from the file).2 In addition, variables that might allow for the reidentification, such as exact date of birth and census tract, have been removed or transposed. The Surveillance, Epidemiology, and End Results Medicare data are not available as public-use files. Investigators who wish to obtain Surveillance, Epidemiology, and End Results Medicare data must sign legally binding data-use agreements with Centers for Medicare and Medicaid Services and Surveillance, Epidemiology, and End Results. In addition, Centers for Medicare and Medicaid Services, NCI, and a representative of the Surveillance, Epidemiology, and End Results registries review all projects before data release to insure there are no concerns about confidentiality. The project was institutional review board–approved and compliant with privacy and guidelines of The Health Insurance Portability and Accountability Act of 1996.
We first selected all Surveillance, Epidemiology, and End Results incident cases of either breast or cervical cancer in women aged 65 years or older diagnosed between 1985 and 2001 that were entitled to Medicare Part A and Part B (Medicare HMO) at the time of diagnosis and who were matched to the Medicare enrollment files. Demographic (age, race, and marital status) and cancer diagnostic information was obtained through Surveillance, Epidemiology, and End Results, and information on Medicare entitlement and use was obtained through Medicare enrollment records. Because of lack of information on individual income and educational level, 1990 U.S. census data were used as proxy measures.
The HMO indicators determined HMO status at the time of diagnosis. If a patient was indicated as not a member of HMO and the claim was processed by Centers for Medicare and Medicaid Services he or she was classified as Medicare fee for service.
For the staging of the breast and cervical cancer, the NCI staging system was used, which is based on the extent of disease at diagnosis as reported by the individual Surveillance, Epidemiology, and End Results registries. These registries abstract information from a variety of sources, including inpatient hospital records, outpatient records, and pathology reports. The histologic staging system consists of 5 tumor stages: in situ, local, regional, distant, and unknown stage. Cancer cases with an unknown stage were excluded from the present analysis.
Patients using a Medicare fee-for-service system may not be using care (because it is not prepaid), so established involvement in a health care delivery system may, in part, be responsible for differences seen in stage of diagnosis for breast and cervical cancer between Medicare HMO and Medicare fee for service. Therefore, we reasoned that patients with a history of cancer would be followed up more closely than patients who do not have cancer. We compared the stage of diagnosis between Medicare HMO and fee for service when another cancer diagnosis preceded the diagnosis of breast or cervical cancer.
Logistic regression analysis was used to evaluate whether HMO status affected the stage at diagnosis for breast or cervical cancer. We used both univariate and multivariate approaches in logistic analysis, with stratification into 4 groups by cancer site and primary and secondary diagnosis as follows: breast cancer as first cancer diagnosis, breast cancer as a second or later cancer diagnosis, cervical cancer as first diagnosis, and cervical cancer as a second or later diagnosis. In the multivariate models, we adjusted for potential confounding factors, including age, race, census tract median income and educational level (proxy measures for socioeconomic status), and marital status. All statistical analyses were performed using the SAS System for Windows V8 (SAS Institute Inc., Cary, NC).
There were a total of 130,336 linked database records for which breast cancer was the first cancer diagnosis. The majority (n = 109,000, 83.6%) was enrolled in Medicare fee for service at time of diagnosis, and 16.4% (n = 21,336) were enrolled in Medicare HMO at the time of cancer diagnosis. Demographic and stage at diagnosis information are shown in Table 1. Fee-for-service patients were slightly older than HMO patients (74.9 years old compared with 74.5 years old, P < .01). The percentage of white patients enrolled in HMO at the time of their cancer diagnosis was significantly lower than those enrolled in fee for service (83.4% compared with 88.5%, P < .01). Compared with fee-for-service, HMO patients were more likely to reside in census tracts with median income greater $20,000 (94.9% compared with 91.3%, P < .01) and higher educational level (> 30% of residents with some college education) (32.6% compared with 29.8, P < .01). The proportion of married women at the time of cancer diagnosis was also higher among HMO patients than fee-for-service patients (46.6% compared with 42.4%, P < .01).
Among patients with breast cancer as their first cancer diagnosis, the proportion of HMO patients diagnosed at stage 0 (in situ, 13.1%) or stage 1 (local, 62.9%) were greater than for fee-for-service patients (10.8% and 59.5%, respectively), whereas patients enrolled in fee for service had greater proportions of regional or later diagnoses (19.8% compared with 24.0% for regional, 4.2% compared with 5.7% for distant). After adjusting for age, race, marital status, and census tract median income and education, logistic regression analysis showed HMO patients were 1.17 times more likely to be diagnosed at stage 0 (in situ, odds ratio [OR] 1.17, 95% confidence interval [CI] 1.12–1.23) and were 1.4 times more likely to be diagnosed at stage 2 or earlier (in situ or local, OR 1.35, 95% CI 1.30–1.40) compared with later-stage diagnoses, as shown in Table 2.
There were a total of 6,933 patients for whom breast cancer was not their first cancer diagnosis. The majority (n = 5,911, 85%) was enrolled in Medicare fee for service at their time of diagnosis. Table 1 shows the demographic and stage at diagnosis for breast cancer when breast cancer was a second or later cancer diagnosis. As compared with fee-for-service patients, HMO patients were slightly younger (72.9 ± 5.9 years old compared with 73.5 ± 6.5 years old, P < .01), less likely to be white, and more likely to reside in a census tract with higher median income (94.8% compared with 90.6%, P < .01), and higher educational level (38.2% compared with 30.8%, P < .01). Medicare HMO patients with breast cancer as their second or higher cancer diagnosis were more likely to be married at the time of their breast cancer diagnosis (43.8% compared with 34.7%) and were more likely to be diagnosed at the in situ stage (14.1% compared with 10.9%). After adjusting for covariates including age, race, marital status, income, and education, patients enrolled in an HMO at the time of their diagnosis did not have a significant increase in diagnoses at stage in situ (OR 1.16, 95% CI 0.94–1.44) as shown in Table 2, either in diagnoses at stages in situ or local stage (OR 1.14, 95% CI 0.95–1.35).
A total of 6,758 patients with cervical cancer as their first cancer diagnosis were identified. Of these, 86.6% of the patients were enrolled in a Medicare fee for service at the time of their cancer diagnosis, and 13.4% were enrolled in an HMO. Demographic information and stage at diagnosis are listed in Table 3. There was no significant difference in the mean age or racial distribution for patients enrolled in an HMO or fee for service. Compared with fee-for-service, HMO patients were significantly more likely to reside in a census tract with a higher median income (87.1% compared with 82.5%, P < .01) and higher educational level (24.8% compared with 19.5%, P < .001), as well as being married at the time of their cancer diagnosis (34.2% compared with 30.7%, P = .04). Health maintenance organization patients had a greater proportion being diagnosed at stage in situ than fee-for-service patients (43.2% compared with 36.5%, P < .01); whereas the proportion being diagnosed at stage local was almost identical between HMO and fee for service (22.8% compared with 22.5%). After adjusting for potential confounding factors, HMO patients (Table 4) were 1.35 times more likely to be diagnosed at stage in situ versus later stages (OR 1.35; 95% CI 1.14–1.59) or to be diagnosed at combined stages in situ and local versus later stages (OR 1.38, 95% CI 1.16–1.64).
There were 61 HMO patients and 498 fee-for-service patients who had cervical cancer as a second or later cancer diagnosis. The mean ages were similar between HMO (73.8 ± 5.9) and fee for service (73.8 ± 6.3). The proportion of whites in HMO was higher than in fee for service (85.3% compared with 78.6%), but the difference was not statistical significant (P = .395). HMO patients were more likely to reside in a census tract with a median income of its residents greater than $20,000 and census tract where the percent of residents with some college education was greater than 30%. The distribution of stage at diagnosis between HMO and fee for service were similar (Table 3). Multivariate logistic regression (Table 4) did not demonstrate any difference for the stage at diagnosis between HMO and fee-for-service patients (in situ compared with later stages, OR 0.91, 95% CI 0.47–1.75; in situ and local compared with later stages, OR 0.89, 95% CI 0.44–1.79) among cervical cancer patients with a prior cancer history.
The main study finding was that Medicare patients enrolled in HMO health care delivery systems were significantly more likely to be diagnosed at an earlier stage for both breast and cervical cancer compared with patients enrolled in a fee-for-service systems. These differences remained even after controlling for potential confounders such as age, race, socioeconomic status, and marital status. The existence of a health care delivery system effect was strengthened by the consistency of our findings for 2 cancer sites that were epidemiologically and clinically quite distinct.
There are several possible explanations for these findings. The earlier diagnosis among HMO patients could have been due to the so-called “HMO effect.” This term has been used to describe the greater likelihood of HMO patients, compared with fee-for-service patients, to use preventive services including disease screening.5 This effect could be related to either plan differences in promotion and access to preventive services or to qualitative differences among HMO patients in terms of education, income, or health consciousness, although the persistence of the effect after adjusting for these factors would indicate some systemic effect. Differences in preventive health services have been found between different health care delivery systems and between uninsured compared with insured populations as well as within racial and ethnic minorities.20,21 For example, Hispanics in HMOs were more likely than their fee-for-service counterparts to report receiving preventive services.4 In a survey of more than 7,500 patients, greater frequency of both clinical breast examination and Pap test use were reported among HMO patients.22
Others have found similar results of differences in the use of screening tests among patients enrolled in different health care delivery systems.4,6,23 In one such study, use of 6 different cancer screening tests (mammography, clinical breast examination, Pap test, fecal occult blood test, and digital rectal examination) varied according to the type of health care coverage. Health maintenance organization enrollees at all ages were approximately 10% more likely to be screened than persons enrolled in private fee-for-service plans.7
Our results are consistent with the findings of Riley et al5 who demonstrated the diagnosis of breast and cervical cancer at an earlier stage over a 5-year period in patients enrolled in HMO compared with fee for service. Interestingly, in that study, although differences were found for some cancers for which screening tests exist (such as breast and cervical), they did not find differences for cancer sites where screening tests are of limited value (such as bladder, uterus, kidney, and ovary). Although we did not study all cancers where available screening tests exist, we examined breast and cervical cancer over a longer time period and attempted to evaluate mechanisms through which these differences occur.
The hypothesis that patients enrolled in HMOs either have greater access to or use of preventive services was supported by our findings that when patients with breast or cervical cancer had been diagnosed with another cancer previously, the observed differences in stage at the time of diagnosis between HMO and fee-for-service systems were no longer present, albeit for cervical cancer the number of observed cases were small. This suggests that the observed differences in stage at diagnosis may be due, at least in part, to an increased use of the 2 health care delivery systems by cancer survivors, rather than systemic differences between the health care delivery systems. Alternatively, the similar findings between the 2 health care delivery systems, for the subpopulation with a history of cancer before their breast or cervical cancer diagnosis, could reflect increased vigilance and screening by patients and providers in both health care delivery plans based on previous cancer history, regardless of health care delivery system. Alternatively, once diagnosed with cancer, patients may educate themselves to the benefit of cancer screening tests or see specialists to a greater extent than those patients without a prior cancer diagnosis. One conclusion is that improving access and vigilance for patients, might to be one way to improve patient outcomes.
In addition to screening use, differences in diagnosis and treatment patterns have been noted between health care delivery systems as well. For earlier stage breast cancer cases requiring surgery, HMO enrollees were significantly more likely to receive radiation therapy.8 Age may also affect treatment choices, because greater use of breast-conserving surgery and radiation therapy has been found among elderly women with early stage breast cancer.24 Both cancer stage at diagnosis and treatment patterns may result in differences in mortality, and possible differences in cancer survival among health care delivery systems should be evaluated as well.
The current study had several limitations. By definition, the study population was aged 65 years or older, and it is not clear that the results are generalizable to a younger patient population. Furthermore, the Surveillance, Epidemiology, and End Results data do not constitute a probability sample of the nation, despite being the primary source of national information on cancer incidence and survival;2,17 Surveillance, Epidemiology, and End Results areas are mostly urban and concentrated in western U.S. states, with undersampling of African Americans. Health plans and patients from these Surveillance, Epidemiology, and End Results areas may not be representative of the nation as a whole. We were not able evaluate all potential confounders such as family history. Additionally, although the study sample was quite large, some strata (particularly among the cervical cancer survivors) had relatively small numbers that may have limited the precision of the statistical findings.
Our findings of earlier stage at diagnosis for both breast and cervical cancer among Medicare HMO patients is important, because early detection of treatable cancers is a fundamental component of cancer prevention and control efforts aimed at decreasing cancer mortality. However, performance among individual or types of HMOs may vary.8 Additionally, other cancer-related patient outcomes are of increasing interest as well. Over the past decade, there has been an increasing demand for the inclusion of patient-centered or reported outcomes that assess the effectiveness of clinical care.25–27 These include measures of symptoms, patient functional status, quality of life, social and emotional consequences of disease, as well as patient satisfaction with care.28,29 Combined with mortality outcomes, measurement of these factors can be used to demonstrate objectively the quality of medical care associated with a given condition.30,31 We recommend further investigation of possible reasons for the differences between health care delivery systems for these outcomes as well.
1. Dutt HR. Developing Medicare HMO market areas and their implications for HMO capitation rates. Manag Care Interface 2003;16:20–5.
2. Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care 2002;40 suppl: IV-3-18.
3. Federman DG, Kirsner RS. The abilities of primary care physicians in dermatology: implications for quality of care. Am J Manag Care 1997;3:1487–92.
4. Delaet DE, Shea S, Carrasquillo O. Receipt of preventative services among privately insured minorities in managed care versus fee-for-service insurance plans. J Gen Intern Med 2002;17:451–7.
5. 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.
6. Potosky AL, Breen N, Graubard BI, Parsons PE. The association between health care coverage and the use of cancer screening tests: results from the 1992 National Health Interview Survey. Med Care 1998;36:257–70.
7. Bernstein AB, Thompson GB, Harlan LC. Differences in rates of cancer screening by usual source of medical care: data from the 1987 National Health Interview Survey. Med Care 1991;29:196–209.
8. 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.
9. American Cancer Society. Cancer facts and figures, 2002. Publication 02-250M-No. 5008.02. Atlanta (GA): American Cancer Society, Inc.; 2002.
10. Fentiman IS. Fixed and modifiable risk factors for breast cancer. Int J Clin Pract 2001;55:527–30.
11. Smith RA, von Eschenbach AC, Wender R, Levin B, Byers T, Rothenberger D, et al. American Cancer Society guidelines for the early detection of cancer: update of early detection guidelines for prostate, colorectal, and endometrial cancers. Also: update 2001—testing for early lung cancer detection [published erratum appears in CA Cancer J Clin 2001;51:150]. CA Cancer J Clin 2001;51:38–75.
12. Brewster A, Helzlsouer K. Breast cancer epidemiology, prevention, and early detection. Curr Opin Oncol 2001;13:420–5.
13. Balducci L, Beghe C. Prevention of cancer in the older person. Clin Geriatr Med 2002;18:505–28.
14. Coronado GD, Thompson B, Koepsell TD, Schwartz SM, McLerran D. Use of Pap test among Hispanics and non-Hispanic whites in a rural setting. Prev Med 2004;38:713–22.
15. Mandelblatt JS, Lawrence WF, Gaffikin L, Limpahayom KK, Lumbiganon P, Warakamin S, et al. Costs and benefits of different strategies to screen for cervical cancer in less-developed countries. J Natl Cancer Inst 2002;94:1469–83.
16. Benard VB, Lee NC, Piper M, Richardson L. Race-specific results of Papanicolaou testing and the rate of cervical neoplasia in the National Breast and Cervical Cancer Early Detection Program, 1991-1998 (United States). Cancer Causes Control 2001;12:61–8.
17. National Cancer Institute. Surveillance, Epidemiology, and End Results. Available at: http://seer.cancer.gov
. Retrieved April 11, 2005.
18. Gornick ME, Warren JL, Eggers PW, Lubitz JD, De Lew N, Davis MH, et al. Thirty years of Medicare: impact on the covered population. Health Care Financ Rev 1996;18:179–237.
19. Potosky AL, Riley GF, Lubitz JD, Mentnech RM, Kessler LG. Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care 1993;31:732–48.
20. Andrulis DP. Access to care is the centerpiece in the elimination of socioeconomic disparities in health. Ann Intern Med 1998;129:412–6.
21. Collins KS, Hall A, Neuhas C. U.S. minorities health: a chartbook. New York (NY): The Commonwealth Fund; 1999.
22. Lonky NM. Reducing death from cervical cancer: examining the prevention paradigms. Obstet Gynecol Clin North Am 2002;29:599–611.
23. Greene J, Blustein J, Laflamme KA. Health Care Financ Rev 2001;22:141–53.
24. Potosky AL, Merrill RM, Riley GF, Taplin SH, Barlow W, Fireman BH, et al. Breast cancer survival and treatment in health maintenance organizations and fee-for-service settings. J Natl Cancer Inst 1997;89:1683–91.
25. Kennedy AW, Austin JM Jr, Look KY, Munger CB. The Society of Gynecologic Oncologists Outcomes Task Force. Study of endometrial cancer: initial experiences. Gynecol Oncol 2000;79:379–98.
26. Blumenthal D. The origins of the quality of care debate. N Engl J Med 1996;335:1146–9.
27. Reemtsma K, Morgan M. Outcomes assessment: a primer. Bull Am Coll Surg 1997;82:34–9.
28. Moinpour CM. Measuring quality of life: an emerging science. Semin Oncol 1994;21:48–60.
29. Testa MA, Simonson DC. Assessment of quality-of-life outcomes. N Engl J Med 1996;334:835–40.
30. Blumenthal D. Part 1: quality of care—what is it? N Engl J Med 1996;335:891–4.
31. Chasin MR. Part 3: improving the quality of care. N Engl J Med 1996;335:1060–3.