Breast Cancer Survival and Treatment in Health
Maintenance Organization and Fee-for-Service
Arnold L. Potosky, Ray M. Merrill, Gerald F. Riley, Stephen H. Taplin,
William Barlow, Bruce H. Fireman, Rachel Ballard-Barbash*
Background: Enrollment in health maintenance organiza-
tions (HMOs) has increased rapidly during the past 10 years,
reflecting a growing emphasis on health care cost contain-
ment. To determine whether there is a difference in the
treatment and outcome for female patients with breast can-
cer enrolled in HMOs versus a fee-for-service setting, we
compared the 10-year survival and initial treatment of pa-
tients with breast cancer enrolled in both types of plans.
Methods: With the use of tumor registries covering the
greater San Francisco–Oakland and Seattle–Puget Sound
areas, respectively, we obtained information on the treat-
ment and outcome for 13 358 female patients with breast
cancer, aged 65 years and older, diagnosed between 1985
and 1992. We linked registry information with Medicare
data and data from the two large HMOs included in the
study. We compared the survival and treatment differences
between HMO and fee-for-service care after adjusting for
tumor stage, comorbidity, and sociodemographic character-
istics. Results: In San Francisco–Oakland, the 10-year ad-
justed risk ratio for breast cancer deaths among HMO pa-
tients compared with fee-for-service patients was 0.71 (95%
confidence interval [CI] = 0.59–0.87) and was comparable
for all deaths. In Seattle–Puget Sound, the risk ratio for
breast cancer deaths was 1.01 (95% CI = 0.77–1.33) but
somewhat lower for all deaths. Women enrolled in HMOs
were more likely to receive breast-conserving surgery than
women in fee-for-service (odds ratio = 1.55 in San Francisco–
Oakland; 3.39 in Seattle). HMO enrollees undergoing
breast-conserving surgery were also more likely to receive
adjuvant radiotherapy (San Francisco–Oakland odds ratio =
2.49; Seattle odds ratio = 4.62). Conclusions: Long-term sur-
vival outcomes in the two prepaid group practice HMOs in
this study were at least equal to, and possibly better than,
outcomes in the fee-for-service system. In addition, the use of
recommended therapy for early stage breast cancer was
more frequent in the two HMOs. [J Natl Cancer Inst 1997;
Breast cancer is the second most common cause of cancer
deaths among women in the United States, trailing only lung
cancer (1). More than 180000 women are newly diagnosed with
breast cancer and 46000 die each year, with about 50% of the
new cases and 80% of the deaths occurring in women ages 65
years and older. There is limited information about breast cancer
care and survival in Health Maintenance Organizations (HMOs).
One study in Orange County, CA, reported poorer survival at 7
years for patients with early stage breast cancer treated in eight
unspecified HMO hospitals compared with those treated in 118
community hospitals, but the analysis did not account for po-
tential confounders, such as patients’ individual insurance status,
socioeconomic characteristics, or comorbidity (2).
From 1980 through 1996, enrollment in HMOs increased
from 4% to 20% of the U.S. population (3), with more than 11%
of Medicare beneficiaries enrolled in HMOs in 1996. While
some HMOs have lower costs and inpatient use, but similar
health outcomes compared with traditional indemnity fee-for-
service (FFS) insurance plans (4–6), there is growing concern
that the rapid growth of the managed care industry may com-
promise the quality of care for those with serious illness (7–9).
Although some believe there is a lack of objective evidence of
any overall decline in the quality of care under the new system
(10), more information is needed on the process and outcomes of
care in managed care settings.
A literature review (5) of managed care plan performance
showed that HMOs have lower hospital admissions yet similar
physician use compared with FFS plans but offer more preven-
tive services. More important, clinical outcome comparisons
were mixed, and most did not adequately adjust for casemix.
Investigators from the Medical Outcomes Study (MOS) found
similar functional outcomes and 7-year mortality rates for pa-
tients with hypertension and diabetes in HMO and FFS settings
(11). In another MOS study, declines in physical health among
the elderly were greater in HMOs than in FFS plans (12). An
evaluation of the Medicare HMO program found similar out-
comes but lower utilization for HMO compared with FFS en-
The purpose of this study was to determine whether there is
a difference in 10-year survival and treatment among women at
least 65 years old diagnosed with breast cancer who were en-
rolled in two large, prepaid-care HMOs and patients in FFS
plans living in the same geographic areas. We also investigated
the influence of other known prognostic variables on survival
*Affiliations of authors: A. L. Potosky, R. M. Merrill, R. Ballard-Barbash,
Applied Research Branch, Division of Cancer Control and Population Sciences,
National Cancer Institute, Bethesda, MD; G. F. Riley, Division of Health Sys-
tems Research at Health Care Financing Administration, Baltimore, MD; S. H.
Taplin, W. Barlow, Center for Health Studies, Group Health Cooperative of
Puget Sound, Seattle, WA; B. H. Fireman, Division of Research, Kaiser Perma-
nente, Northern California Region, Oakland, CA.
Correspondence to: Arnold L. Potosky, Ph.D., National Institutes of Health,
Executive Plaza North, Rm. 313, Bethesda, MD 20892–7344. E-mail:
See ‘‘Notes’’ following ‘‘References.’’
© Oxford University Press
Journal of the National Cancer Institute, Vol. 89, No. 22, November 19, 1997 ARTICLES 1683
by guest on December 28, 2015
including age, race, tumor stage, socioeconomic status (SES),
and presence of comorbidity. We compared the use of treatments
for early stage breast cancer cases in HMO and FFS settings.
Data Sources and Sample Selection
Data for breast cancer cases diagnosed from 1985 through 1992 were obtained
from two population-based cancer registries, one covering the five-county San
Francisco–Oakland greater metropolitan area and the other covering the 13-
county Seattle–Puget Sound area. Both registries, which cover at least half of the
catchment areas of the two HMOs included in the study, also participate in the
National Cancer Institute’s (NCI) Surveillance, Epidemiology, and End Results1
(SEER) Program. The SEER Program consists of a system of 11 population-
based tumor registries covering approximately 14% of the United States (1).
Each registry collects and reports information on all new cancer diagnoses,
including date of diagnosis, stage at diagnosis, whether the cancer is a single or
multiple primary tumor, and the date and cause of death. Stage of disease at
initial diagnosis for breast cancer cases is identified using the TNM (tumor–
node–metastasis) classification system adopted by the American Joint Commit-
tee on Cancer (15). Another prognostic variable collected by SEER, estrogen
receptor status, was not included, since this item was not routinely collected until
the early 1990s. For all cases, we used information collected by the SEER
registries on the type of initial surgery and radiation therapy administered or
planned during the 4 months following diagnosis. Vital status follow-up at the
time of our study was complete through December 1994.
We used a SEER-Medicare linked database to augment the SEER registry
data. The methods used and the results of linking SEER with Medicare data are
described elsewhere (16). More than 94% of all SEER cases aged 65 years and
older were successfully linked with Medicare claims data. Because our study
required information from the SEER–Medicare linkage, eligible cases for the
study were women diagnosed with breast cancer at ages 65 years and older
whose SEER record had been previously linked with Medicare files. We in-
cluded women with breast cancer as either a single primary diagnosis or as the
first primary diagnosis (with at least one or more subsequent new primary tumors
of any site including breast). Newly diagnosed breast cancer cases diagnosed
between 1985 and 1992 in San Francisco–Oakland and Seattle–Puget Sound
areas were followed for a maximum of 10 years through December 1994. The
median follow-up time was 52 months.
We identified all new breast cancer cases reported to the two SEER registries
by the two largest HMO plans in these regions, Kaiser Permanente of Northern
California (KPNC) and Group Health Cooperative of Puget Sound (GHC).
KPNC, the largest and one of the oldest prepaid group plans, was established in
the early 1940s and during 1987 through 1990 covered approximately 2.4 mil-
lion members in Northern California, of which about half reside in the five
counties covered by the SEER registry. GHC was established as the first con-
sumer-controlled health plan in 1947. Unlike KPNC, which contracts with a
distinct physician organization (The Permanente Medical Group) to provide
services to the plan members, GHC directly employs salaried physicians. GHC
covered about 385000 persons in the Puget Sound area during the study period.
Each HMO submitted patient lists to NCI containing the unique SEER case
identification number for any patient diagnosed between 1985 and 1992 as
having breast cancer while enrolled in the HMO. Personal identifiers were not
required, since SEER cases are tracked by case identification numbers that are
not associated with names or social security numbers. More than 91% of these
HMO patients were continuously enrolled in their HMO from diagnosis through
death or last date of follow-up. All cases were classified according to their
enrollment status at the time of diagnosis and initial treatment for breast cancer
and remained in the HMO category regardless of disenrollment.
Inpatient Medicare claims data spanning 1984 through 1992 for FFS breast
cancer cases diagnosed from 1985 through 1992 were used to identify the pres-
ence of comorbid conditions at the time of initial diagnosis. To measure comor-
bidity, we used a Charlson/Deyo comorbidity index. This index was designed
and validated using Medicare inpatient hospital claims containing ICD-9-CM
diagnostic codes (17,18). The modified Charlson/Deyo Index is not a measure of
all comorbid diseases, but rather a risk measure for mortality from other causes
than the reference condition in our study, breast cancer. Since Medicare inpatient
claims data are incomplete for HMOs paid under capitation, we obtained auto-
mated inpatient data directly from KPNC and GHC internal accounting systems.
Internal audits of HMO automated inpatient data show that the accuracy of
diagnoses recorded is quite high. Inpatient data covering 1984 through 1992
were linked to individual HMO cases by use of the unique SEER case ID
number. The HMO inpatient data contained similar information to the Medicare
FFS inpatient claims data, including dates of admission and discharge and up to
five ICD-9-CM diagnostic and three surgical procedure codes. We used infor-
mation from inpatient records from the 2 years before diagnosis to compute all
the Charlson/Deyo comorbidity index scores reported in this study.
We obtained group-level income and educational attainment variables by
linking the census tract reported to SEER for each case with group variables
collected by the U.S. Bureau of the Census during the 1990 decennial census.
The variables used were the median household income of the census tract and the
proportion of adults with less than a high school education. For approximately
8% of the cases, census tracts were unavailable so ZIP Code level information
We began with a sample of 14171 female breast cancer cases linked with
Medicare claims data who were diagnosed in the two regions between 1985 and
1992. We excluded patients enrolled in any other HMO besides KPNC or GHC
(n ? 780), since we lacked information on HMO type and comorbidity status,
and another 33 cases with missing group education and income values, leaving
a final sample size of 13358 (7176 in San Francisco–Oakland; 6182 in Seattle).
Frequency distributions of age, race, income, education, tumor stage, and
comorbidity were compared with enrollment status separately by region. We
used standard categorical chi-squared tests to evaluate statistical differences
between HMO and FFS cases in the distributions of each independent variable.
For patients diagnosed from 1985 through 1992 and followed through De-
cember 1994, we calculated 10-year Kaplan–Meier survival probabilities for
deaths from all causes, testing for differences between HMO and FFS settings
within each region. Comparisons were stratified by age and stage, but not both
simultaneously. We assessed differences between HMO and FFS Kaplan–Meier
survival functions using the logrank test.
The Cox proportional hazards regression model was used to estimate the
relative risk of overall and breast cancer-specific mortality in the HMO breast
cancer cases compared with those in the Medicare FFS system. The Cox regres-
sion models permitted estimation of the effects of HMO enrollment status on
mortality risk, after adjustment for other patient characteristics, including age at
diagnosis, race, stage at diagnosis, whether the diagnosis was a single or first
primary cancer, comorbidity, and area-level educational status. We constructed
separate survival models to assess overall and breast cancer-specific mortality. In
the latter models, we censored nonbreast cancer deaths. We included only one
measure of socioeconomic status (SES) because median income and educational
status are highly correlated. We included categoric variables for age, race, edu-
cation (quartiles), stage at diagnosis, single or first primary cancer, and comor-
bidity. The use of continuous rather than categoric variables for age and educa-
tion level did not affect the results.
In the Cox models, we tested for two-way interaction terms between HMO
enrollment status and age, stage, comorbidity, and region. We detected only one
significant interaction term, enrollment status by region, so we constructed sepa-
rate survival models for Seattle and San Francisco–Oakland. Standard log–log
plots were used to evaluate each covariate with respect to the validity of the
proportional hazards assumption. No evidence of substantial departures from this
assumption was observed.
We constructed separate logistic regression models to estimate the odds ratios
for two different treatments comparing HMO to FFS cases with in situ or stage
I or II tumors, controlling for all other independent variables contained in the
survival models. In one regression model, we examined the use of breast-
conserving surgery (BCS) versus mastectomy, and in the second model, the use
of adjuvant radiotherapy following surgery among all women undergoing BCS.
We defined mastectomy using SEER surgical coding as modified radical mas-
tectomy, simple mastectomy, subcutaneous mastectomy, or radical mastectomy.
BCS was defined as any of the following: segmental resection, excisional biopsy,
lumpectomy, wedge resection, quadrantectomy, tylectomy, nipple resection, or
Table 1 shows frequency distributions of selected patient
characteristics in each area. A higher percentage of HMO cases
1684 ARTICLES Journal of the National Cancer Institute, Vol. 89, No. 22, November 19, 1997
by guest on December 28, 2015
mammographic screening for breast cancer. Am J Epidemiol 1994;140:
(35) 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;
(36) Campbell KM, Holm K. Preventive service utilization among older
women: a comparison of HMO members, private insurance policy holders,
and Medicare recipients. Olympia (WA): Washington State Department of
Health, Center for Health Statistics, 1995.
(37) Fisher B, Redmond C, Poisson R, Margolese R, Wolmark N, Wickerham
L, et al. Eight-year results of a randomized clinical trial comparing total
mastectomy and lumpectomy with or without irradiation in the treatment of
breast cancer [published erratum appears in N Engl J Med 1994;330:1467].
N Engl J Med 1989;320:822–8.
(38) Clark RM, Whelan T, Levine M, Roberts R, Willan A, McCulloch P, et al.
Randomized clinical trial of breast irradiation following lumpectomy and
axillary dissection for node-negative breast cancer: an update. J Natl Can-
cer Inst 1996;88:1659–64.
(39) Veronesi U, Luini A, Del Vecchio M, Greco M, Galimberti V, Merson M,
et al. Radiotherapy after breast-preserving surgery in women with localized
cancer of the breast. N Engl J Med 1993;328:1587–91.
(40) NIH consensus conference. Treatment of early-stage breast cancer. JAMA
(41) Baue AE. Breast-conservation operations for treatment of cancer of the
breast [editorial]. JAMA 1994;271:1204–5.
(42) Riley G, Tudor C, Chiang YP, Ingber M. Health status of Medicare en-
rollees in HMOs and fee-for-service in 1994. Health Care Financ Rev
(43) Retchin SM, Clement DG, Rossiter LF, Brown B, Brown R, Nelson L.
How the elderly fare in HMOs: outcomes from the Medicare competition
demonstrations. Health Serv Res 1992;27:651–69.
(44) Pappas G, Queen S, Hadden W, Fisher G. The increasing disparity in
mortality between socioeconomic groups in the United States, 1960 and
1986 [published erratum appears in N Engl J Med 1993;329:1139]. N Engl
J Med 1993;329:103–9.
(45) Krieger N. Overcoming the absence of socioeconomic data in medical
records: validation and application of a census-based methodology. Am J
Public Health 1992;82:703–10.
(46) Dayal HH, Power RN, Chiu C. Race and socio-economic status in survival
from breast cancer. J Chronic Dis 1982;35:675–83.
(47) Greenwald HP, Polissar NL, Borgatta EF, McCorkle R. Detecting survival
effects of socioeconomic status: problems in the use of aggregate measures.
J Clin Epidemiol 1994;47:903–9.
1Editor’s note: SEER is a set of geographically defined, population-based
central tumor registries in the United States, operated by local nonprofit orga-
nizations under contract to the National Cancer Institute (NCI). Each registry
annually submits its cases to the NCI on a computer tape. These computer tapes
are then edited by the NCI and made available for analysis.
2From Health Plan Employer Data Information Set report, 1995.
We thank Jim Rovan, Information Management Services, Inc., for his out-
standing data processing assistance, and James D. Lubitz, Health Care Financing
Administration, for his helpful comments on the manuscript.
Manuscript received April 7, 1997; revised August 19, 1997; accepted Sep-
tember 11, 1997.
Journal of the National Cancer Institute, Vol. 89, No. 22, November 19, 1997ARTICLES 1691
by guest on December 28, 2015