ArticlePDF Available

Contribution Of Care Source To Cancer Treatment Cost Variation In The US Military Health System

Authors:

Abstract

The US Military Health System (MHS) provides universal access to health care for more than nine million eligible beneficiaries through direct care in military treatment facilities or purchased care in civilian facilities. Using information from linked cancer registry and administrative databases, we examined how care source contributed to cancer treatment cost variation in the MHS for patients ages 18-64 who were diagnosed with colon, female breast, or prostate cancer in the period 2003-14. After accounting for patient, tumor, and treatment characteristics, we found the independent contribution of care source to total variation in cost to be 8 percent, 12 percent, and 2 percent for colon, breast, and prostate cancer treatment, respectively. About 20-50 percent of the total cost variance remained unexplained and may be related to organizational and administrative factors.
By Yvonne L. Eaglehouse, Mayada Aljehani, Matthew W. Georg, Olga Castellanos, Jerry S. H. Lee,
Seth A. Seabury, Craig D. Shriver, and Kangmin Zhu
Contribution Of Care Source To
Cancer Treatment Cost Variation In
The US Military Health System
ABSTRACT
The US Military Health System (MHS) provides universal access
to health care for more than nine million eligible beneficiaries through
direct care in military treatment facilities or purchased care in civilian
facilities. Using information from linked cancer registry and
administrative databases, we examined how care source contributed to
cancer treatment cost variation in the MHS for patients ages 1864 who
were diagnosed with colon, female breast, or prostate cancer in the
period 200314. After accounting for patient, tumor, and treatment
characteristics, we found the independent contribution of care source to
total variation in cost to be 8 percent, 12 percent, and 2 percent for
colon, breast, and prostate cancer treatment, respectively. About 20
50 percent of the total cost variance remained unexplained and may be
related to organizational and administrative factors.
Cancer care imposes a significant fi-
nancial burden on US health sys-
tems, with an estimated total annual
cost of $150 billion in 2018.1Based
on current cancer incidence and the
growing number of cancer survivors, as well as
the rapid development of new treatments, the
cost for treating and managing cancer is ex-
pected to rise over the next several decades. In
the US Military Health System (MHS) approxi-
mately 7,000 new primary malignant cancers are
diagnosed each year among Department of De-
fense (DoD) beneficiaries, including active duty
service members, National Guard members, re-
tirees, and their eligible family members and
dependents.24The annual cost for treating can-
cer within the MHS was estimated to exceed
$1 billion in 2002, which is the most recent esti-
mate available.3The current annual costs to the
MHS are likely to be higher, owing to the trends
in cancer incidence and the development of
newer treatmentswhich have contributed to
increased costs in the United States as a whole.5
The MHS provides health care services to more
than nine million eligible beneficiaries through
direct care in military treatment facilities or pur-
chased care in the civilian sector. The Defense
Health Agency (DHA) provides operational and
administrative oversight for shared services,
functions, and activities in the MHS (for exam-
ple, TRICARE health plan, health information
technology, research, and acquisition) that sup-
port the Army, Navy, Air Force, and Marine
Corps.6,7 A unified medical budget appropriated
to the DoD provides financial support for the
DHA and MHS activities.6The DHA allocates
funding for direct care at military treatment fa-
cilities based partly on the number of eligible
beneficiaries in the military treatment facility
catchment area, patient volume, and types of
services offered at the facility.8The cost to the
facility for each patient visit is determined using
patient-level cost allocation methods. The direct
medical expenses at the facility are allocated
down to the patient-encounter level. The allocat-
ed costs do not include the operational, admin-
istrative, or overhead costs at the facility. The
DHA also allocates funds to manage payments
for purchased care.6Costs to the MHS for medi-
cal services at civilian facilities are determined
doi: 10.1377/hlthaff.2019.00283
HEALTH AFFAIRS 38,
NO. 8 (2019): 13351342
©2019 Project HOPE
The People-to-People Health
Foundation, Inc.
Yvonne L. Eaglehouse is a
health services researcher in
the Murtha Cancer Center
Research Program,
Department of Surgery,
Uniformed Services University
of the Health Sciences
(USUHS); is an assistant
professor in the Department
of Surgery at USUHS; and is
employed by the Henry M.
Jackson Foundation for the
Advancement of Military
Medicine, all in Bethesda,
Maryland.
Mayada Aljehani is a
biostatistician in the
Lawrence J. Ellison Institute
for Transformative Medicine,
University of Southern
California (USC), in Los
Angeles.
Matthew W. Georg is a
research associate in the
Murtha Cancer Center
Research Program,
Department of Surgery,
USUHS; and is employed by
the Henry M. Jackson
Foundation for the
Advancement of Military
Medicine.
Olga Castellanos is a clinical
research program manager in
theLawrenceJ.Ellison
Institute for Transformative
Medicine, USC.
Jerry S. H. Lee is the chief
science and innovation officer
intheLawrenceJ.Ellison
Institute for Transformative
Medicine,USC;isanassociate
professor in the Departments
of Clinical Medicine and
Chemical Engineering, both at
USC; and is employed by the
Henry M. Jackson Foundation
for the Advancement of
Military Medicine.
August 2019 38:8 Health Affairs 1335
Military Health System
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
from standard fee schedules9and depend on pro-
vider network status and the cost-sharing ar-
rangements of the patients benefit plan.8,10
The DoD provides insurance coverage to
beneficiaries through TRICARE.10 The primary
TRICARE benefit plan options include Prime,
Select (formerly called Extra and Standard),
and TRICARE for Life.8TRICARE for Life serves
as a secondary payer to Medicare for people who
are eligible for and enrolled in Medicare. Bene-
ficiaries with TRICARE pay nothing out of pocket
when they receive care at military treatment fa-
cilities. Beneficiaries may have copayments or
point-of-service fees due at the time of service
when they receive purchased care, depending on
their benefit plan, the providers network status,
and whether out-of-pocket maximums have been
reached. Payments for specialty treatment (for
example, for cancer) in purchased care are han-
dled in a way similar to that in which a health
maintenance organization (in the case of Prime)
or preferred provider organization (in the case of
Select) in the private sector handles them.6,10
Although TRICARE plans are similar in structure
to these private managed care models, MHS ben-
eficiaries generally have lower out-of-pocket ex-
penses than do people who participate in US
public or private plans.8
Previous studies have investigated care source
in relation to cancer screening, treatment, sur-
veillance, or clinical outcomes in the MHS.1115 A
few studies have compared costs for cancer treat-
ment paid by the DoD between care sources.1618
Two of these found lower per patient costs for
breast and colon cancer treatment in direct care
compared to costs for treatment in purchased
care,16,17 while the third found similar costs for
head and neck cancer treatment in the care
sources.18 The contributors to observed cost var-
iation in cancer treatment in the MHS are un-
known. The purpose of the present study was to
examine the contributions of care source to can-
cer treatment cost variation in the MHS, inde-
pendent of patient, tumor, and treatment fac-
tors, for three common cancer sites: colon,
female breast, and prostate.
Study Data And Methods
Population And Inclusion Criteria This study
was a retrospective analysis of linked data from
the DoDs Central Cancer Registry19 and the MHS
Data Repository medical claims database.20 The
registry contains information on patients diag-
nosed with or treated for cancer in military treat-
ment facilities that is consolidated following
the criteria of the North American Association
of Central Cancer Registries. Patients with path-
ologically confirmed invasive colon, female
breast, or prostate cancer were identified in
the Central Cancer Registry data using topogra-
phy codes from the International Classification of
DiseasesOncology, Third Edition (ICD-O-3).21
Eligible subjects were patients ages 1864 who
were diagnosed with a single primary colon, fe-
male breast, or prostate cancer in the period
January 1, 2003December 31, 2014 (see online
appendix exhibit A1 for information about our
selection of patients for the study).22 We exclud-
ed patients ages sixty-five and older at diagnosis
because of possibly incomplete medical claims
and cost data for services covered under Medi-
care, since TRICARE is a secondary payer to
Medicare. Patients with other insurance (in ad-
dition to MHS-provided coverage) were also ex-
cluded because of the possibility of missing treat-
ment and medical claims and cost data.
The original data linkage project was reviewed
and approved by the Institutional Review Boards
of the Walter Reed National Military Medical
Center and the Defense Health Agency.
Study Variables Cancer diagnosis informa-
tion included pathologic and clinical tumor
stage, consolidated into the American Joint
Committee on Cancers stages I, II, III, and
IV;23,24 tumor grade, defined by the committee
as well differentiated (G1), moderately differen-
tiated (G2), poorly differentiated (G3), undiffer-
entiated (G4), or unknown/undetermined
(GX);23,24 tumor size; and cancer sitespecific
variables (for example, prostate specific antigen
levels for prostate cancer, estrogen receptor and
progesterone receptor status for breast cancer).
Demographic characteristics included sex, race,
ethnicity, and age and marital status at diagno-
sis. Military-specific characteristics included
active-duty status at diagnosis, military service
or sponsor branch, and TRICARE service region.
Comorbidities were identified from MHS Data
Repository records using International Classifica-
tion of Diseases, Ninth Revision (ICD-9), codes
for conditions included in the Charlson Comor-
bidity Index.25
Care source was determined from information
in the MHS Data Repository for cancer treatment
by methods described previously.16,17 Briefly,
unique claim records for cancer treatment deliv-
ered in direct or purchased care in the first
year after cancer diagnosis were extracted, and
visit counts were summed per treatment type
(see appendix exhibit A2 for a list of codes).22
The relative proportion of claims for direct care
were used to determine where patients received
each treatment type (that is, surgery, chemother-
apy, radiotherapy, and hormone therapy). Then
the proportion of total treatment types received
in direct care were used to determine the care
source. If at least 80 percent of treatment types
Seth A. Seabury is the
director of the Keck-Schaeffer
Initiative for Population
Health Policy at the Leonard
D. Schaeffer Center for Health
Policy and Economics and an
associate professor in the
Department of Pharmaceutical
and Health Economics at the
School of Pharmacy, both at
USC.
Craig D. Shriver is the
director of the Murtha
Cancer Center Research
Program, Department of
Surgery, USUHS; director of
the Murtha Cancer Center at
Walter Reed National Military
Medical Center; and a
professor in the Department
of Surgery at USUHS.
Kangmin Zhu (kzhu@
murthacancercenter.org) is the
director of Military
Epidemiology and Population
Science in the Murtha Cancer
Center Research Program,
Department of Surgery,
USUHS; is a professor in the
Department of Preventive
Medicine and Biostatistics at
USUHS; and is employed by
the Henry M. Jackson
Foundation for the
Advancement of Military
Medicine.
Military Health System
1336 Health Affairs August 2019 38:8
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
were received as direct care, then the care source
was classified as direct. If 20 percent or less of
treatment types were received as direct care, then
the care source was classified as purchased. Oth-
erwise, patients who used a mix of direct and
purchased care (2080 percent of treatment
types in each) were classified as using both care
sources. Patients who did not have cancer treat-
ment (and therefore had no associated costs)
documented in the MHS Data Repository rec-
ords during the study period were excluded (ap-
pendix exhibit A1).22
Outcome The costs to the MHS (paid by the
DoD) for cancer treatment in the first year after
diagnosis were extracted from the MHS Data
Repository for inpatient and outpatient services
received in direct or purchased care. The costs
did not include patient-paid costs such as enroll-
ment fees and copayments. The first year after
diagnosis was selected as the study period be-
cause that is when primary treatment is indicat-
ed by recommended guidelines,2628 and other
studies have shown that cancer care expendi-
tures are highest in this time period.1,29,30 Rele-
vant encounters were identified using specific
diagnostic and treatment codes used in the
ICD-9 orafter October 1, 2015the Internation-
al Statistical Classification of Diseases and Related
Health Problems, Tenth Revision (ICD-10); the
Healthcare Common Procedure Coding System;
and Current Procedural Terminology (appendix ex-
hibit A2).22 For treatment in military treatment
facilities (direct care), the total costs recorded in
the data represent episodes of care including
professional and outpatient ancillary costs (for
example, laboratory, radiology, and pharmacy
costs) related to a procedure (such as surgery).
For treatment in civilian facilities (purchased
care), the costs in the data include the amount
paid by the DoD for each billed procedure but
none of the associated professional and ancillary
costs. To increase cost comparability between
direct and purchased care, costs for clinician
salary were subtracted from the total direct care
costs related to cancer treatment for inpatient
claims, and costs for ancillary services were sub-
tracted from the total direct care costs for cancer
treatment when not relevant for outpatient
claims. Then the costs for cancer treatment were
summed per patient and adjusted to 2018 dol-
lars, using the medical component of the Con-
sumer Price Index.31
Statistical Analysis We used box plots to
examine the distribution between care sources
of the cost for cancer treatment in the three tu-
mor sites.We used generalized linear models to
estimate the mean cost and associated standard
errors per patient for cancer treatment within
each care source for the first year after diagnosis.
We used sequential hierarchical generalized lin-
ear mixed models with gamma distribution and
log link function32,33 to estimate the amount of
cost variation attributed to care sources and to
patient, military, tumor, and treatment charac-
teristics.We started with a model that contained
only the intercept as a random effect and tested
the univariable effects of each potential contrib-
utor as fixed effects.We began sequential model-
ing with a base model that contained the inter-
cept and the variable for care source, then we
added patient, tumor, and treatment variables
as fixed effects to the base model to determine
the total amount of variance explained by the
factors. The sequential models included adjust-
ment for variables in the previous nested model.
The residual covariance parameter estimates for
each model were used to calculate the reduction
in outcome (cost) variance for the nested mod-
els. Last, we estimated the amount of total varia-
tion in cost attributed to care source, indepen-
dent of other variables, by calculating the
difference in the covariance parameter estimates
between the full model and the full model minus
care source as a portion of the total model vari-
ance from the intercept model. Hierarchical gen-
eralized linear mixed models have previously
been used to study expenditures in prostate
and breast cancer.34,35
Li mitat io ns Our study had several limita-
tions. First, we assessed direct medical costs to
the MHS for cancer treatment. These did not
include administrative, operational, or overhead
costs. Thus, total costs to the MHS for cancer
care were likely higher than estimated, and the
results should be interpreted in the context of
treatment costs.
Second, the comparability of direct and pur-
chased care costs was limited by differences be-
tween the two in the way the data are recorded;
incentives and rules for coding; bill creation; and
cost allocation and budgeting. However, we took
steps to improve the comparability of cost esti-
mates, as explained above.
Third, care source was defined based on the
percentage of visits related to cancer-specific
care that were received as direct care. The selec-
tion of the 80 percent cut point could have re-
sulted in misclassification of costs in the analysis
(that is, some costs of direct care might have
been included in the costs of purchased care,
and vice versa). Also, patients with multiple
treatment types were more likely to use both care
sources based on our definition, and this could
have affected our interpretation of the results.
Finally, errors in the administrative data could
have resulted in the inaccurate capture of cancer
treatment and costs. These errors were infre-
quent and might not have affected the results.
August 2019 38:8 Health Affairs 1337
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
Study Results
There were 15,166 patients who were diagnosed
and treated for colon (n=1,629), female breast
(n=6,945), or prostate (n=6,592) cancer in the
period January 1, 2003December 31, 2014 (ap-
pendix exhibit A1).22 Patientsdemographic and
tumor characteristics are shown in exhibit 1 for
each cancer site, and the corresponding infor-
mation by care source is in appendix exhibits A3
A5.22 In comparing distributions of characteris-
tics within each cancer site, we found that pa-
tients who used direct care were more likely to be
male (those with colon cancer), on active duty
(breast and prostate cancer), or of minority race/
ethnicity than patients who used purchased care
or both sources. Patients with breast or colon
cancer who used purchased care tended to be
younger and non-Hispanic black. Patients with
these cancers who used direct care were more
likely to have stage I or II or G1 tumors, compared
to those who used purchased care or both sourc-
es. Patients with prostate cancer who used direct
care tended to have more stage III or IV tumors,
but more favorable Gleason scores (a prostate
cancerspecific factor), compared to patients us-
ing purchased care.
Treatment for each tumor site by care source is
shown in exhibit 2. In general, patients who used
direct care were equally or more likely to have
surgery than those who used purchased care.
Chemotherapy use was typically higher among
patients who used purchased care. Radiotherapy
use was generally higher among patients who
used both care sources. Hormone treatment
use was higher among patients with breast can-
cer, but lower among patients with prostate can-
cer, who used direct care, compared to those who
used purchased care or both care sources.
For colon cancer, patients who used direct care
had the lowest median cost ($41,568), followed
by those who used purchased care ($125,647)
(exhibit 3). Those who used both care sources
had the highest median treatment cost
($173,469). For breast cancer, patients who used
direct care had the lowest median treatment cost
($37,889), while those who used purchased care
($57,054) or both care sources ($57,074) had
higher median cost. There was no significant
median cost difference between the use of pur-
chased care or both care sources. For prostate
cancer, patients who used purchased care had
the lowest median treatment cost ($13,104), fol-
lowed by those who used direct care ($17,220)
and those who used both care sources ($18,914).
There were also significant differences between
care sources in unadjusted estimated mean costs
(exhibit 3). The higher mean values compared to
medians reflect the presence of outliers, partic-
ularly among patients who used purchased care
and both care sources. Treatment cost and cost
variability were highest for colon cancer and low-
est for prostate cancer.
In univariable generalized linear mixed mod-
els, the estimated total variance in cost attribut-
ed to care source was 37 percent in both colon
and prostate cancer and 29 percent in breast
cancer (exhibit 4, model 0). In addition, tumor
stage and chemotherapy treatment (for colon
Exhibit 1
Demographic and tumor characteristics of patients diagnosed with colon, female breast, or
prostate cancer in the US Military Health System, 200314
Colon
(n=1,629)
Breast
(n=6,945)
Prostate
(n=6,592)
Age at diagnosis (years)
1839 13.3% 15.2% 0.4%
4049 20.9 33.2 16.0
5059 41.4 33.1 47.7
6064 24.4 18.5 35.9
Race/ethnicity
Non-Hispanic white 53.4 53.1 51.0
Non-Hispanic black 20.3 16.9 25.7
Non-Hispanic Asian/PI/AIAN 9.7 12.5 5.1
Hispanic 6.9 6.6 5.0
Unknown 9.6 11.0 13.1
Marital status at diagnosis
Married 79.7 84.7 81.9
Single 6.1 4.1 4.4
Divorced/widowed/separated 11.1 9.4 8.5
Unknown 3.0 1.8 5.2
Military service or sponsor branch
Army 31.8 33.8 33.9
Navy 21.2 19.9 18.3
Marine Corps 3.8 4.7 4.0
Air Force 24.8 26.6 27.5
Other 7.9 5.5 7.1
Unknown 10.4 9.5 9.1
On active duty at diagnosis
Yes 17.6 8.7 15.8
No 72.6 82.3 75.6
Unknown 9.9 9.0 8.6
Tumor stagea
I 23.4 44.7 10.7
II 19.9 36.3 74.4
III 30.2 13.7 8.4
IV 23.2 3.2 3.5
Unknown 3.3 2.1 2.8
Tumor gradea,b
G1 17.6 19.6 7.3
G2 57.8 36.2 55.7
G3 13.6 34.0 31.2
G4 0.9 0.7 0.5
GX 10.2 9.4 5.3
SOURCE Authorsanalysis of data for 200314 from the Department of Defense Central Cancer
Registry and Military Health System Data Repository linked databases. NOTES Appendix
exhibits A3A5 provide more details and information by care source (see note 22 in text). PI is
Pacific Islander. AIAN is American Indian or Alaska Native. aFor cases diagnosed before 2010, per
GreeneFL,etal.,editors.AJCCcancerstagingmanual.6thed.p.421(note23intext).Forcases
diagnosed in or after 2010, per Edge S, et al., editors, AJCC cancer staging manual. 7th ed. p. 648
(note 24 in text). bTumor grades are explained in the text.
Military Health System
1338 Health Affairs August 2019 38:8
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
and breast cancer), hormone treatment and TRI-
CARE region (breast and prostate), and surgery
and radiation (prostate) accounted for substan-
tial amounts of cost variation in univariable
models (data not shown). In sequential hierar-
chical generalized linear mixed models, we esti-
mated the variance in cost accounted for by the
addition of patient, military, tumor, and treat-
ment characteristics (models 14 in exhibit 4).
The contribution of each added set of explanato-
ry variables yielded different results across the
three tumor sites (appendix exhibit A7).22 In to-
tal, the explanatory variables accounted for
61 percent of the cost variation in colon cancer,
52 percent in breast cancer, and 81 percent in
prostate cancer (model 4 in exhibit 4). When we
compared the model that adjusted for patient,
tumor, and treatment characteristics to the full
model that also adjusted for care source, we
found that the total model variation in cost at-
tributed to care source independent of other var-
iables was 8 percent in colon cancer, 12 percent
in breast cancer, and 2 percent in prostate cancer
(data not shown). We repeated the analysis ex-
cluding patients who died or whose data were
censored within the first year after diagnosis to
reduce the potential effects on total cost and cost
differences, and we found similar results (data
not shown).
Discussion
In the Military Health System, treatment costs
per patient with colon, female breast, or prostate
cancer varied significantly between care sources,
with lower median and mean costs in colon and
breast cancer and lower mean costs in prostate
cancer for patients who used direct care than
for those who used purchased care or both care
sources. This is consistent with two previous
studies of breast and colon cancer in the
MHS.16,17 In the current study, care source ex-
plained about 3040 percent of observed varia-
tion in cost per patient before any other factors
were adjusted for. The addition of patient, tu-
mor, and treatment characteristics accounted
for another 2045 percent of the total variation
in cancer care costs. After these other factors
were adjusted for, the amount of total variance
in cost to the MHS explained by care source was
8 percent in colon cancer, 12 percent in breast
cancer, and 2 percent in prostate cancer. The
remaining unexplained model variance in cost
was 39 percent in colon cancer, 48 percent in
breast cancer, and 19 percent in prostate cancer.
Care source made a significant independent
contribution to variations in costs for cancer
treatment, although we accounted for potential
differences in patient, tumor, and treatment
characteristics. Health system structure and oth-
er unmeasured factors, such as market supply
and demand and institution type, may also con-
tribute to variability in cost between care sourc-
es. Being able to accurately decompose institu-
tional factors such as guideline adherence and
prices, as well as heterogeneity in physician prac-
tice patterns, from patient and disease character-
istics will be critical to designing strategies to
optimize value in cancer care, both within the
MHS and more broadly in the civilian sector.
Tumor and treatment characteristics ac-
Exhibit 2
Cancer treatment for patients diagnosed with colon, female breast, or prostate cancer in the
US Military Health System, by care source, 200314
Direct care Purchased care Both
Colon cancer
Number of patients 974 311 344
Surgery
Yes 93.4% 80.7% 95.9%
No 6.6 19.3 4.1
Chemotherapy
Yes 35.4 70.4 93.3
No 64.6 29.6 6.7
Radiotherapy
Yes 14.8 14.1 35.2
No 85.2 85.8 64.8
Female breast cancer
Number of patients 2,217 2,437 2,291
Surgery
Yes 89.5% 93.1% 98.4%
No 10.5 6.8 1.6
Chemotherapy
Yes 50.7 63.4 76.2
No 49.3 36.6 23.7
Radiotherapy
Yes 62.1 57.2 83.1
No 37.9 42.8 16.9
Hormone
Yes 46.6 22.8 35.6
No 31.8 56.8 41.0
Not applicablea21.6 20.3 23.4
Prostate cancer
Number of patients 2,741 2,325 1,526
Surgery
Yes 76.6% 42.6% 78.4%
No 23.4 57.4 21.6
Chemotherapy
Yes 6.4 17.3 19.5
No 93.6 82.7 80.5
Radiotherapy
Yes 25.8 35.1 33.5
No 74.2 64.9 66.5
Hormone
Yes 6.6 91.0 94.9
No 93.4 9.0 5.1
SOURCE Authorsanalysis of data for 200314 from the Department of Defense Central Cancer
Registry and Military Health System Data Repository linked databases. aPatients with hormone
receptor negative breast cancer (estrogen receptor negative or progesterone receptor negative).
August 2019 38:8 Health Affairs 1339
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
counted for a large proportion of cost variability.
In post hoc analysis we examined the potential
contribution of care source to cost variance in
subgroups of patients with similar characteris-
tics. In colon and breast cancer, care source ac-
counted for a significant amount of cost varia-
tion among patients who received surgery and
chemotherapy, regardless of tumor stage (data
not shown). In prostate cancer, there was more
variability in cost for radiotherapy associated
with care source in the West compared to other
regions (data not shown). While these analyses
provide some clues to cost differences within
more homogeneous populations, the interpreta-
tion was limited by the smaller sample sizes and
residual variability within each patient group.
Variation in case-mix and treatment patterns be-
tween care sources, as shown in our data, may be
important for further explaining cost differenc-
es. However, other studies of breast and colon
cancer costs in the MHS have found differences
in cost between care sources even within treat-
ment types.16,17 Further investigation of the use of
specific treatment modalities that may vary in
cost and cost-effectiveness3638 and whether they
are differentially administered between direct
and purchased care is needed to better under-
stand cost variability.
The landscape of cancer treatment is changing
quickly, with the recent approval by the Food and
Drug Administration (FDA)39 and national cov-
erage by the Centers for Medicare and Medicaid
Services (CMS)40 of next-generation sequencing
tests and the FDAs approval of two tumor type
agnostic, biomarker-driven treatments.41 Given
the rapid adoption of novel treatments and pre-
cision medicine initiatives,42,43 it will be critical
to understand the interplay between tumor and
treatment characteristics and variance in cost. In
the MHS, patients with more advanced or com-
plex disease may have been referred to pur-
chased care when the local military treatment
facility did not have treatment capabilities, thus
affecting the cost for care. In some tumor sites,
tumor pathology and other clinical features may
largely determine the type of treatment required
and whether patients receive a referral for pur-
chased care. Patientspreferences for treatment
may also drive the choice of care sources. With
the diffusion of innovative technologies into
community practice, these cancer- and patient-
specific characteristics are likely to become an
important determinant of cancer care cost as
oncologists and other providers move toward
personalized cancer treatments that are typically
more expensive than conventional therapies.42,44
As the MHS continues to transition to organi-
zational management under the DHA,7in which
the DHA will provide oversight for both direct
and purchased care, research such as that de-
scribed here will become more important for
guiding decision making in resource allocation
and provision of care. In the MHS, as with the
civilian sector, not all facilities have the capabili-
ty to provide oncology services, and it might not
be feasible for all to do so. While this study is
preliminary and more in-depth analysis of other
factors that drive cost (for example, market sup-
ply and demand, institutional status and type,
and treatment use) is required to inform such
provisional changes in services, we have provid-
ed some evidence that cancer treatment costs
might be lower with direct care and that the
variability in these costs might be controlled.
It is still unclear whether the potential benefit
of increasing service capabilities at military
treatment facilities to provide cancer treatment
that would otherwise be received in purchased
care would outweigh the administrative and op-
erational costs.
It should be noted that our study compared
differences in treatment costs between care
sources, but it did not assess differences in the
Exhibit 3
Treatment cost per patient in the first year after diagnosis for patients with colon, female
breast, or prostate cancer using direct care, purchased care, or both, 200314
SOURCE Authorsanalysis of data for 200314 from the Department of Defense Central Cancer Reg-
istry and the Military Health System Data Repository l inked databases. NOTES The exhibit shows box
plot distributions of cost in 2018 dollars. The horizontal line within each box plot represents the
median cost. The twenty-fifth and seventy-fifth percentiles (interquartile ranges) are represented
by the height of the boxes, variability outside the upper and lower quartiles is represented by
the whiskers, and unadjusted means are represented by diamonds. Complete box plot distributions,
including data outliers, are in appendix exhibit A6 (see note 22 in text). For colon cancer and prostate
cancer, median costs were significantly different between care sources (p<0:001 for all pairwise
comparisons). For breast cancer, median costs in purchased care and both care sources were signifi-
cantly different from direct care (p<0:001) but not from each other (p¼0:37). For all three tumor
sites, the unadjusted mean costs were significantly different between care sources (p<0:001 for all
pairwise comparisons).
Military Health System
1340 Health Affairs August 2019 38:8
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
value of care.45,46 This would require more com-
prehensive and specific clinical outcome data
such as cancer as the cause of death and infor-
mation about care-seeking behavior, patients
preferences for direct or purchased care, and
patient satisfactionall of which were unavail-
able for the current study. If the costs per patient
in purchased care represented simply higher
prices or greater utilization with no improve-
ment in outcomes, it would suggest a need for
cost containment strategies.
Conclusion
In the MHS, cancer treatment costs per patient
in direct care were generally lower than costs per
patient in purchased care or both sources for
colon, female breast, and prostate cancer. Pa-
tient, tumor, and treatment characteristics ac-
counted for some, but not all, of the total vari-
ance in cancer treatment costs. Future efforts are
needed to examine the specific treatment com-
ponents that drive cost and to explore organiza-
tional and institutional factors that may contrib-
ute to variations in spending for cancer care.
This project was supported by the
Murtha Cancer Center Research
Program, Department of Surgery,
Uniformed Services University of the
Health Sciences (USUHS), and the
Walter Reed National Military Medical
Center under the auspices of the Henry
M. Jackson Foundation for the
Advancement of Military Medicine. The
contents of this publication are the sole
responsibility of the authors and do not
reflect the views, assertions, opinions,
or policies of the USUHS; the
Department of Defense; the
Departments of the Army, Navy, or Air
Force; or any other agency of the US
government. Mention of trade names,
commercial products, or organizations
does not imply endorsement by the US
government.
NOTES
1National Cancer Institute. Cancer
trends progress report: financial
burden of cancer care [Internet].
Bethesda (MD): National Cancer
Institute; 2019 Feb [cited 2019 May
30]. Available from: https://
progressreport.cancer.gov/after/
economic_burden
2Lee T, Williams VF, Clark LL. Inci-
dent diagnoses of cancers in the ac-
tive component and cancer-related
deaths in the active and reserve
components, U.S. Armed Forces,
20052014. MSMR. 2016;23(7):
2331.
3Crawford RS 3rd, Wu J, Park D,
Barbour GL. A study of cancer in the
military beneficiary population. Mil
Med. 2007;172(10):10848.
4Zhu K, Devesa SS, Wu H, Zahm SH,
Jatoi I, Anderson WF, et al. Cancer
incidence in the U.S. military popu-
lation: comparison with rates from
the SEER program. Cancer Epide-
miol Biomarkers Prev. 2009;18(6):
17405.
5Dieleman JL, Squires E, Bui AL,
Campbell M, Chapin A, Hamavid H,
et al. Factors associated with in-
creases in US health care spending,
19962013. JAMA. 2017;318(17):
166878.
6Jansen DJ. Military medical care:
questions and answers [Internet].
Washington (DC): Congressional
Research Service; 2014 Jan 2 [cited
2019 May 30]. Available from:
https://fas.org/sgp/crs/misc/
RL33537.pdf
7Health.mil. Defense Health Agency
[home page on the Internet]. Falls
Church (VA): DHA; [cited 2019 May
30]. Available from: https://www
.health.mil/About-MHS/OASDHA/
Defense-Health-Agency
8Health.mil. Evaluation of the TRI-
CARE Program: fiscal year 2019 re-
port to Congress [Internet]. Falls
Church (VA): Defense Health Agen-
cy; 2019 Feb 28 [cited 2019 Jul 10].
Available for download from:
https://www.health.mil/Military-
Health-Topics/Access-Cost-Quality-
and-Safety/Health-Care-Program-
Evaluation/Annual-Evaluation-of-
the-TRICARE-Program
9Health.mil. Rates and reimburse-
ment [Internet]. Falls Church (VA):
Defense Health Agency; 2017 [cited
2019 May 29]. Available from:
https://www.health.mil/military-
health-topics/business-support/
rates-and-reimbursement
10 Tricare.mil. TRICARE [home page
on the Internet]. Falls Church (VA):
Defense Health Agency; [cited 2019
May 30]. Available from: https://
tricare.mil/
11 Manjelievskaia J, Brown D, Shao S,
Hofmann K, Shriver CD, Zhu K.
Benefit type and care source in re-
lation to mammography screening
Exhibit 4
Percent of variance in cancer treatment cost in the US Military Health System explained by
linear mixed effects models, 200314
SOURCE Authorsanalysis of data for 200314 from the Department of Defense Central Cancer Reg-
istry and the Military Health System Data Repository linked databases. NOTES The models present
results by hierarchical generalized linear mixed effects. Model 0 is intercept plus care source. Model
1 adds to model 0 age, sex, race/ethnicity, marital status, and comorbid conditions. Model 2 adds to
model 1 active-duty status, military service or sponsor branch, and TRICARE service region. Model 3
adds to model 2 tumor stage, grade, histology, diagnosis year, subsite or location, and other site-
specific variables (for example, prostate specific antigen levels for prostate cancer or estrogen re-
ceptor and progesterone receptor status for breast cancer). Model 4 adds to model 3 treatment
type (surgery, chemotherapy, radiotherapy, or hormone) and treatment volume (number of visits
or encounters).
August 2019 38:8 Health Affairs 1341
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
and breast cancer stage at diagnosis
among DoD beneficiaries. Mil Med.
2017;182(3):e17829.
12 Manjelievskaia J, Brown D, Shao S,
Hofmann K, Shriver CD, Zhu K.
Breast cancer treatment and survival
among Department of Defense ben-
eficiaries: an analysis by benefit type
and care source. Mil Med. 2018;
183(34):e18695.
13 Chaudhary MA, Leow JJ, Mossanen
M, Chowdhury R, Jiang W, Learn PA,
et al. Patient driven care in the
management of prostate cancer:
analysis of the United States Military
Healthcare System. BMC Urol. 2017;
17(1):56.
14 Fox JP, Jeffery DD, Williams TV,
Gross CP. Quality of cancer survi-
vorship care in the Military Health
System (TRICARE). Cancer J. 2013;
19(1):19.
15 Eaglehouse YL, Shao S, Chan W,
Brown D, Manjelievskaia J, Shriver
CD, et al. The continuum of breast
cancer care and outcomes in the U.S.
Military Health System: an analysis
by benefit type and care source. J
Cancer Surviv. 2018;12(3):40716.
16 Eaglehouse YL, Manjelievskaia J,
Shao S, Brown D, Hofmann K,
Richard P, et al. Costs for breast
cancer care in the Military Health
System: an analysis by benefit type
and care source. Mil Med. 2018;
183(1112):e5008.
17 Eaglehouse YL, Georg MW, Richard
P, Shriver CCD, Zhu K. Costs for
colon cancer treatment comparing
benefit types and care sources in the
US Military Health System. Mil Med.
2019 Apr 3. [Epub ahead of print].
18 Ambrosio A, Jeffery DD, Hopkins L,
Burke HB. Cost and healthcare uti-
lization among non-elderly head and
neck cancer patients in the Military
Health System, a single-payer uni-
versal health care model. Mil Med.
2019;184(56):e4007.
19 Joint Pathology Center. DoD cancer
registry program [Internet]. Silver
Spring (MD): JPC; [last modified
2014 Nov 4; cited 2019 May 30].
Available from: https://www.jpc
.capmed.mil/education/dodccrs/
index.asp l
20 Health.mil. Military Health System
Data Repository [Internet]. Falls
Church (VA): Defense Health Agen-
cy; [cited 2019 May 30]. Available
from: https://www.health.mil/
Military-Health-Topics/Technology/
Clinical-Support/Military-Health-
System-Data-Repository
21 International Agency for Research
on Cancer. International Classifica-
tion of Diseases for Oncology: ICD-
O-3 online [Internet]. Lyon
(France): IARC; c 2019 [cited 2019
May 30]. Available from: http://
codes.iarc.fr
22 To access the appendix, click on the
Details tab of the article online.
23 Greene FL, Page DL, Fleming ID,
Fritz A, Balch CM, Haller DG, edi-
tors. AJCC cancer staging manual.
6th ed. New York (NY): Springer-
Verlag; 2002. p. 421.
24 Edge S, Byrd DR, Compton CC, Fritz
AG, Greene FL, Trotti A, editors.
AJCC cancer staging manual. 7th ed.
New York (NY): Springer-Verlag;
2010. p. 648.
25 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(5):37383.
26 Benson AB 3rd, Venook AP,
Cederquist L, Chan E, Chen YJ,
Cooper HS, et al. Colon cancer, ver-
sion 1.2017, NCCN clinical practice
guidelines in oncology. J Natl Compr
Canc Netw. 2017;15(3):37098.
27 Gradishar WJ, Anderson BO,
Balassanian R, Blair SL, Burstein HJ,
Cyr A, et al. Breast cancer, version
4.2017, NCCN clinical practice
guidelines in oncology. J Natl Compr
Canc Netw. 2018;16(3):31020.
28 Mohler JL, Armstrong AJ, Bahnson
RR, DAmico AV, Davis BJ, Eastham
JA, et al. Prostate cancer, version
1.2016. J Natl Compr Canc Netw.
2016;14(1):1930.
29 Brown ML, Riley GF, Schussler N,
Etzioni R. Estimating health care
costs related to cancer treatment
from SEER-Medicare data. Med
Care. 2002;40(8, Suppl):IV10417.
30 Warren JL, Yabroff KR, Meekins A,
Topor M, Lamont EB, Brown ML.
Evaluation of trends in the cost of
initial cancer treatment. J Natl Can-
cer Inst. 2008;100(12):88897.
31 Bureau of Labor Statistics. CPIAll
Urban Consumers (Current Series)
[Internet]. Washington (DC): BLS;
[last updated 2018 Nov 22; cited
2019 June 10]. Available from:
https://catalog.data.gov/dataset/
consumer-price-index-all-urban-
consumers-current-series
32 Lininger M, Spybrook J, Cheatham
CC. Hierarchical linear model:
thinking outside the traditional
repeated-measures analysis-of-vari-
ance box. J Athl Train. 2015;50(4):
43841.
33 Lee Y, Nelder JA. Hierarchical gen-
eralized linear models. J R Stat Soc
Series B Stat Methodol. 1996;58(4):
61978.
34 Wang SY,Wang R, Yu JB, Ma X, Xu X,
Kim SP, et al. Understanding re-
gional variation in Medicare expen-
ditures for initial episodes of pros-
tate cancer care. Med Care. 2014;
52(8):6807.
35 Xu X, Herrin J, Soulos PR, Saraf A,
Roberts KB, Killelea BK, et al. The
role of patient factors, cancer char-
acteristics, and treatment patterns in
the cost of care for Medicare bene-
ficiaries with breast cancer. Health
Serv Res. 2016;51(1):16786.
36 Soni A, Chu E. Cost-effectiveness of
adjuvant chemotherapy in the treat-
ment of early-stage colon cancer.
Clin Colorectal Cancer. 2015;14(4):
21926.
37 Deshmukh AA, Shirvani SM, Lal L,
Swint JM, Cantor SB, Smith BD,
et al. Cost-effectiveness analysis
comparing conventional, hypofrac-
tionated, and intraoperative radio-
therapy for early-stage breast cancer.
J Natl Cancer Inst. 2017;109(11).
38 Schroeck FR, Jacobs BL, Bhayani SB,
Nguyen PL, Penson D, Hu J. Cost of
new technologies in prostate cancer
treatment: systematic review of costs
and cost effectiveness of robotic-
assisted laparoscopic prostatectomy,
intensity-modulated radiotherapy,
and proton beam therapy. Eur Urol.
2017;72(5):71235.
39 Food and Drug Administration [In-
ternet]. Silver Spring (MD): FDA;
2017. Press release, FDA announces
approval, CMS proposes coverage of
first breakthrough-designated test to
detect extensive number of cancer
biomarkers; 2017 Nov 30 [cited 2019
May 30]. Available from: https://
www.fda.gov/news-events/press-
announcements/fda-announces-
approval-cms-proposes-coverage-
first-breakthrough-designated-test-
detect-extensive
40 CMS.gov [Internet]. Baltimore
(MD): Centers for Medicare and
Medicaid Services; 2018. Press re-
lease, CMS finalizes coverage of Next
Generation Sequencing tests, en-
suring enhanced access for cancer
patients; 2018 Mar 16 [cited 2019
May 30]. Available from: https://
www.cms.gov/newsroom/press-
releases/
41 Yan L, Zhang W. Precision medicine
becomes realitytumor typeagnos-
tic therapy. Cancer Commun. 2018;
38(6):17.
42 Prasad V. Immunotherapy:
tisagenlecleucelthe first approved
CAR-T-cell therapy: implications for
payers and policy makers. Nat Rev
Clin Oncol. 2018;15(1):112.
43 Wilson AL, Plebanski M, Stephens
AN. New trends in anti-cancer ther-
apy: combining conventional che-
motherapeutics with novel immu-
nomodulators. Curr Med Chem.
2018;25(36):475884.
44 Schnipper LE, Meropol N. ASCO
addresses the rising cost of cancer
care. J Oncol Pract. 2009;5(4):
2145.
45 Shih YC, Ganz PA, Aberle D,
Abernethy A, Bekelman J, Brawley
O, et al. Delivering high-quality and
affordable care throughout the can-
cer care continuum. J Clin Oncol.
2013;31(32):41517.
46 Schnipper LE, Davidson NE, Wollins
DS, Tyne C, Blayney DW, Blum D,
et al. American Society of Clinical
Oncology statement: a conceptual
framework to assess the value of
cancer treatment options. J Clin
Oncol. 2015;33(23):256377.
Military Health System
1342 Health Affairs August 2019 38:8
Downloaded from HealthAffairs.org on August 05, 2019.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Precision medicine just witnessed two breakthroughs in oncology in 2017. Pembrolizumab (Keytruda), Merck’s anti-programmed cell death-1 (PD-1) monoclonal antibody (mAb), received accelerated approval in May 2017 by the US Food and Drug Administration for the treatment of adult and pediatric patients with unresectable or metastatic solid tumors that have been identified as having microsatellite instability-high (MSI-H) or deficient DNA mismatch repair (dMMR). Shortly after, nivolumab (Opdivo), Bristol-Myers Squibb’s anti-PD-1 mAb, gained an accelerated approval in August 2017 for adult and pediatric patients with MSI-H or dMMR metastatic colorectal cancer that has progressed after standard chemotherapy. These regulatory approvals marked an important milestone that a cancer treatment may be approved based on a common biomarker rather than the anatomic location in the body where the tumor originated, and therefore established a precedent for tumor type-agnostic therapy. In the 2017 American Society for Clinical Oncology annual meeting, larotrectinib (LOXO-101), Loxooncology’s oral, potent, and selective inhibitor of tropomyosin receptor kinases (TRK), demonstrated unprecedented efficacy on unresectable or metastatic solid tumors with neurotrophic tropomyosin receptor kinase (NTRK)-fusion proteins in adult and pediatric patients. Both the anti-PD-1 mAbs and the TRK-targeting therapies share some basic features: (a) biomarker-based, well-defined rare patient population; (b) exceptionally high clinical efficacy, e.g., near 40% overall response rate (ORR) for pembrolizumab across 15 tumor types with MSI-H/dMMR and 75% ORR for larotrectinib across more than 12 tumor types with NTRK-fusion proteins; (c) durable responses lasting at least 6 months with complete responses observed; and (d) parallel development in adult and pediatric populations. With increasing accessibility to genetic analysis tools such as next-generation sequencing, tumor type-agnostic therapy has become a reality, both during clinical development and in clinical practice. Adjustments in our approaches to developing new anti-cancer drugs and to adopting these new cancer treatments in clinical practice need to occur in order to prepare ourselves for the new era of precision medicine.
Article
Full-text available
Ductal carcinoma in situ (DCIS) of the breast represents a heterogeneous group of neoplastic lesions in the breast ducts. The goal for management of DCIS is to prevent the development of invasive breast cancer. This manuscript focuses on the NCCN Guidelines Panel recommendations for the workup, primary treatment, risk reduction strategies, and surveillance specific to DCIS.
Article
Full-text available
Purpose: This study investigates transition rates between breast cancer diagnosis, recurrence, and death by insurance benefit type and care source in U.S. Military Health System (MHS). Methods: The MHS data repository and central cancer registry linked data were used to identify women aged 40-64 with histologically confirmed breast cancer between 2003 and 2007. Three-state continuous time Markov models were used to estimate transition rates and transition rate ratios (TRRs) by TRICARE benefit type (Prime or non-Prime) and care source (direct, purchased, or both), adjusted for demographic, tumor, and treatment variables. Results: Analyses included 2668 women with transitions from diagnosis to recurrence (n = 832), recurrence to death (n = 79), and diagnosis to death without recurrence (n = 91). Compared to women with Prime within each care source, women with non-Prime using both care sources had higher transition rates (TRR 1.47, 95% CI 1.03, 2.10). Compared to those using direct care within each benefit type, women utilizing both care sources with non-Prime had higher transition rates (TRR 1.86, 95% CI 1.11, 3.13), while women with Prime utilizing purchased care had lower transition rates (TRR 0.82, 95% CI 0.68, 0.98). Conclusions: In the MHS, women with non-Prime benefit plans compared to Prime had higher transition rates along the breast cancer continuum among both care source users. Purchased care users had lower transition rates than direct care users among Prime beneficiaries. Implications for cancer survivors: Benefit plan and care source may be associated with breast cancer progression. Further research is needed to demonstrate differences in survivorship.
Article
Full-text available
Importance Health care spending in the United States increased substantially from 1995 to 2015 and comprised 17.8% of the economy in 2015. Understanding the relationship between known factors and spending increases over time could inform policy efforts to contain future spending growth. Objective To quantify changes in spending associated with 5 fundamental factors related to health care spending in the United States: population size, population age structure, disease prevalence or incidence, service utilization, and service price and intensity. Design and Setting Data on the 5 factors from 1996 through 2013 were extracted for 155 health conditions, 36 age and sex groups, and 6 types of care from the Global Burden of Disease 2015 study and the Institute for Health Metrics and Evaluation’s US Disease Expenditure 2013 project. Decomposition analysis was performed to estimate the association between changes in these factors and changes in health care spending and to estimate the variability across health conditions and types of care. Exposures Change in population size, population aging, disease prevalence or incidence, service utilization, or service price and intensity. Main Outcomes and Measures Change in health care spending from 1996 through 2013. Results After adjustments for price inflation, annual health care spending on inpatient, ambulatory, retail pharmaceutical, nursing facility, emergency department, and dental care increased by $933.5 billion between 1996 and 2013, from $1.2 trillion to $2.1 trillion. Increases in US population size were associated with a 23.1% (uncertainty interval [UI], 23.1%-23.1%), or $269.5 (UI, $269.0-$270.0) billion, spending increase; aging of the population was associated with an 11.6% (UI, 11.4%-11.8%), or $135.7 (UI, $133.3-$137.7) billion, spending increase. Changes in disease prevalence or incidence were associated with spending reductions of 2.4% (UI, 0.9%-3.8%), or $28.2 (UI, $10.5-$44.4) billion, whereas changes in service utilization were not associated with a statistically significant change in spending. Changes in service price and intensity were associated with a 50.0% (UI, 45.0%-55.0%), or $583.5 (UI, $525.2-$641.4) billion, spending increase. The influence of these 5 factors varied by health condition and type of care. For example, the increase in annual diabetes spending between 1996 and 2013 was $64.4 (UI, $57.9-$70.6) billion; $44.4 (UI, $38.7-$49.6) billion of this increase was pharmaceutical spending. Conclusions and Relevance Increases in US health care spending from 1996 through 2013 were largely related to increases in health care service price and intensity but were also positively associated with population growth and aging and negatively associated with disease prevalence or incidence. Understanding these factors and their variability across health conditions and types of care may inform policy efforts to contain health care spending.
Article
Introduction: Cancer is one of the leading causes of morbidity and mortality in the USA, contributing largely to US healthcare spending. Provision of services (direct or purchased) and insurance benefit type may impact cost for cancer care. As a common cause of cancer in both men and women, we aim to compare colon cancer treatment costs between insurance benefit types and care sources in the US Military Health System (MHS) to better understand whether and to what extent these system factors impact cancer care costs. Materials and methods: Department of Defense Central Cancer Registry records and MHS Data Repository administrative claims were used to identify MHS beneficiaries aged 18-64 who were diagnosed with primary colon adenocarcinoma and received treatment between 2003 and 2008. The data linkage was approved by the Institutional Review Boards of the Walter Reed National Military Medical Center, the Defense Health Agency, and the National Institutes of Health. Costs to the MHS for each claim related to cancer treatment were extracted from the linked data and adjusted to 2008 USD. We used quantile regression models to compare median cancer treatment costs between benefit types and care sources (direct, purchased, or both), adjusted for demographic, tumor, and treatment characteristics. Results: The median per capita (n = 801) costs for colon cancer care were $60,321 (interquartile range $24,625, $159,729) over a median follow-up of 1.7 years. The model-estimated treatment costs were similar between benefit types. Patients using direct care had significantly lower estimated median costs [$34,145 (standard error $4,326)] than patients using purchased care [$106,395 ($10,559)] or both care sources [$82,439 ($13,330)], controlled for patient demographic, tumor, and treatment characteristics. Differences in cost by care source were noted for patients with later stage tumors and by treatment type. Relative costs were 2-3 times higher for purchased care compared to direct care for patients with late-stage tumors and for patients receiving chemotherapy or radiation treatment. Conclusions: In the MHS, median cost for colon cancer treatment was lower in direct care compared to purchased care or patients using a combination of direct and purchased care. The variation in cancer treatment costs between care sources may be due to differences in treatment incentives or capabilities. Additional studies on cost differences between direct and purchased services are needed to understand how provision of care affects cancer treatment costs and to identify possible targets for cost reduction.
Article
We consider hierarchical generalized linear models which allow extra error components in the linear predictors of generalized linear models. The distribution of these components is not restricted to be normal; this allows a broader class of models, which includes generalized linear mixed models. We use a generalization of Henderson's joint likelihood, called a hierarchical or h‐likelihood, for inferences from hierarchical generalized linear models. This avoids the integration that is necessary when marginal likelihood is used. Under appropriate conditions maximizing the h‐likelihood gives fixed effect estimators that are asymptotically equivalent to those obtained from the use of marginal likelihood; at the same time we obtain the random effect estimates that are asymptotically best unbiased predictors. An adjusted profile h‐likelihood is shown to give the required generalization of restricted maximum likelihood for the estimation of dispersion components. A scaled deviance test for the goodness of fit, a model selection criterion for choosing between various dispersion models and a graphical method for checking the distributional assumption of random effects are proposed. The ideas of quasi‐likelihood and extended quasi‐likelihood are generalized to the new class. We give examples of the Poisson–gamma, binomial–beta and gamma–inverse gamma hierarchical generalized linear models. A resolution is proposed for the apparent difference between population‐averaged and subject‐specific models. A unified framework is provided for viewing and extending many existing methods.
Article
Introduction Examining costs and utilization in a single-payer universal health care system provides information on fiscal and resource burdens associated with head and neck cancer (HNC). Here, we examine trends in the Department of Defense (DoD) HNC population with respect to: (1) reimbursed annual costs and (2) patterns and predictors of health care utilization in military only, civilian only, and both systems of care (mixed model). Materials and Methods A retrospective, cross-sectional study was conducted using TRICARE claims data from fiscal years 2007 through 2014 for reimbursement of ambulatory, inpatient, and pharmacy charges. The study was approved by the Defense Health Agency Office of Privacy and Civil Liberties as exempt from institutional review board full review. The population was all beneficiaries, age 18–64, with a primary ICD-9 diagnosis of HNC, on average, 2,944 HNC cases per year. The outcomes of regression models were total reimbursed health care cost, and counts of ambulatory visits, hospitalizations, and bed days. The predictors were fiscal year, demographic variables, hospice use, type and geographic region of TRICARE enrollment, use of military or civilian care or mixed use, cancer treatment modalities, the number of physical and mental health comorbid conditions, and tobacco use. A priori, null hypotheses were assumed. Results Per annual average, 61% of the HNC population was age 55–64, and 69% were males. About 6% accessed military facilities only for all health care, 60% accessed civilian only, and 34% accessed both military and civilian facilities. Patients who only accessed military care had earlier stage disease as indicated by rates of single modality treatment and hospice use; military care only and mixed use had similar rates of combination treatment and hospice use. The average cost per patient per year was $14,050 for civilian care only, $13,036 for military care only, and $29,338 for mixed use of both systems. The strongest predictors of higher cost were chemotherapy, radiation therapy, head and neck surgery, hospice care, and mixed-use care. The strongest predictors of health care utilization were chemotherapy, use of hospice, the number of physical and mental health comorbidities, radiation therapy, head and neck surgery, and system of care. Conclusions To a single payer, the use of a single system of care exclusively among HNC patients is more cost-effective than use of a mixed-use system. The results suggest an over-utilization of ambulatory care services when both military and civilian care are accessed. Further investigation is needed to assess coordination between systems of care and improved efficiencies with respect to the cost and apparent over-utilization of health care services.
Article
Introduction Breast cancer care imposes a significant financial burden to U.S. healthcare systems. Health services factors, such as insurance benefit type and care source, may impact costs to the health system. Beneficiaries in the U.S. Military Health System (MHS) have universal healthcare coverage and access to a network of military facilities (direct care) and private practices (purchased care). This study aims to quantify and compare breast cancer care costs to the MHS by insurance benefit type and care source. Materials and Methods We conducted a retrospective analysis of data linked between the MHS data repository administrative claims and central cancer registry databases. The institutional review boards of the Walter Reed National Military Medical Center, the Defense Health Agency, and the National Institutes of Health Office of Human Subjects Research reviewed and approved the data linkage. We used the linked data to identify records for women aged 40–64 yr who were diagnosed with breast cancer between 2003 and 2007 and to extract information on insurance benefit type, care source, and cost to the MHS for breast cancer treatment. We estimated per capita costs for breast cancer care by benefit type and care source in 2008 USD using generalized linear models, adjusted for demographic, pathologic, and treatment characteristics. Results The average per capita (n = 2,666) total cost for breast cancer care was $66,300 [standard error (SE) $9,200] over 3.31 (1.48) years of follow-up. Total costs were similar between benefit types, but varied by care source. The average per capita cost was $34,500 ($3,000) for direct care (n = 924), $96,800 ($4,800) for purchased care (n = 622), and $60,700 ($3,900) for both care sources (n = 1,120), respectively. Care source differences remained by tumor stage and for chemotherapy, radiation, and hormone therapy treatment types. Conclusions Per capita costs to the MHS for breast cancer care were similar by benefit type and lower for direct care compared with purchased care. Further research is needed in breast and other tumor sites to determine patterns and determinants of cancer care costs between benefit types and care sources within the MHS.
Article
Background: Use of treatment for breast cancer is dependent on the patient's cancer characteristics and willingness to undergo treatment and provider treatment recommendations. Receipt of breast cancer treatment varies by insurance status and type. It is not clear whether different benefit types and care sources differ in breast cancer. Methods: The objectives of this study are to assess whether receipt of breast cancer treatment varied by benefit type (TRICARE Prime vs non-Prime) or care source (direct care, purchased care, and both) and to examine whether survival and recurrence differed by benefit type and/or care source among female Department of Defense beneficiaries with the disease. Study subjects were women aged 40-64 yr, diagnosed with malignant breast cancer between 2003 and 2007. Multivariable logistic regression analyses were conducted to assess the likelihood of receiving treatment by benefit type or care source. Multivariable Cox proportional hazard models were used to investigate differences in survival and recurrence by benefit type or care source. Findings: A total of 2,668 women were included in this study. Those with Prime were more likely to have chemotherapy, radiation, hormone therapy, breast-conserving surgery, surveillance mammography, and recurrence than women with non-Prime. Survival was high, with 94.86% of those with Prime and 92.58% with non-Prime alive at the end of the study period. Women aged 50-59 yr with non-Prime benefit type had better survival than women with Prime of the same age. No survival differences were seen by care source. In regard to recurrence, women aged 60-64 yr with TRICARE Prime were more likely to have recurrent breast cancer than women with non-Prime. Additionally, women aged 50-59 yr who used purchased care were less likely to have a recurrence than women who used direct care only. Discussion/impact/recommendations: To our knowledge, this is the first study to examine breast cancer treatment and survival by care source and benefit type in the Military Health System. In this equal access health care system, no differences in treatment, except mastectomy, by benefit type, were observed. There were no overall differences in survival, although patients with non-Prime tended to have better survival in the age group of 50-59 yr. In regard to care source, women who utilized mostly purchased care or utilized both direct and purchased care were more likely to receive certain types of treatment, such as chemotherapy and radiation, as compared with women who used direct care only. However, survival did not differ between different care sources. Future research is warranted to further investigate variations in breast cancer treatment and its survival gains by benefit type and care source among Department of Defense beneficiaries.
Book
The American Joint Committee on Cancer's Cancer Staging Manual is used by physicians throughout the world to diagnose cancer and determine the extent to which cancer has progressed. All of the TNM staging information included in this Sixth Edition is uniform between the AJCC (American Joint Committee on Cancer) and the UICC (International Union Against Cancer). In addition to the information found in the Handbook, the Manual provides standardized data forms for each anatomic site, which can be utilized as permanent patient records, enabling clinicians and cancer research scientists to maintain consistency in evaluating the efficacy of diagnosis and treatment. The CD-ROM packaged with each Manual contains printable copies of each of the book’s 45 Staging Forms.