Fax: (770) 488-4639
Has the burden shifted over time?
Article first published online: 10 MAY 2010
Copyright © 2010 American Cancer Society
Volume 116, Issue 14, pages 3477–3484, 15 July 2010
How to Cite
Tangka, F. K., Trogdon, J. G., Richardson, L. C., Howard, D., Sabatino, S. A. and Finkelstein, E. A. (2010), Cancer treatment cost in the United States. Cancer, 116: 3477–3484. doi: 10.1002/cncr.25150
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
This article is United States government work and, as such, is in the public domain in the United States of America
- Issue published online: 2 JUL 2010
- Article first published online: 10 MAY 2010
- Manuscript Accepted: 10 NOV 2009
- Manuscript Revised: 2 NOV 2009
- Manuscript Received: 10 SEP 2009
- medical expenditures;
- health insurance;
- cost of illness;
- Medical Expenditure Panel Survey;
- National Medical Expenditure Survey
There has not been a comprehensive analysis of how aggregate cancer costs have changed over time. The authors present 1) updated estimates of the prevalence and total cost of cancer for select payers and how these have changed over the past 2 decades; and 2) for each payer, the distribution of payments by type of service over time to assess whether there have been shifts in cancer treatment settings.
Pooled data from the 2001 through 2005 Medical Expenditure Panel Survey and the 1987 National Medical Care Expenditure Survey were used for the analysis. The authors used an econometric approach to estimate cancer-attributable medical expenditures by payer and type of service.
In 1987, the total medical cost of cancer (in 2007 US dollars) was $24.7 billion. Private payers financed the largest share of the total (42%), followed by Medicare (33%), out of pocket (17%), other public (7%), and Medicaid (1%). Between 1987 and the 2001 to 2005 period, the total medical cost of cancer increased to $48.1 billion. In 2001 to 2005, the shares of cancer costs were: private insurance (50%), Medicare (34%), out of pocket (8%), other public (5%), and Medicaid (3%). The share of total cancer costs that resulted from inpatient admissions fell from 64.4% in 1987 to 27.5% in 2001 to 2005.
The authors identified 3 trends in the total costs of cancer: 1) the medical costs of cancer have nearly doubled; 2) cancer costs have shifted away from the inpatient setting; and 3) the share of these costs paid for by private insurance and Medicaid have increased. Cancer 2010. Published 2010 by American Cancer Society.
A solid understanding of trends in the level and composition of spending on cancer, in conjunction with data on outcomes, is important for evaluating whether we are spending our healthcare resources wisely and for prioritizing future resources for treating and preventing cancer. Total spending on cancer and the composition of spending is likely to vary over time for several reasons. First, over the past several decades, cancer-related incidence and prevalence has changed significantly. Whereas the annual percentage increase in cancer incidence was 1.2% per year in the late 1970s and 1980s, incidence decreased by roughly 1.7% per year during 2001 to 2005.1 Furthermore, prevalence, defined as the number or persons previously diagnosed with cancer, has increased over time. In 2004, >10 million Americans were living with cancer, including roughly 1.4 million who were diagnosed within the past 12 months.2 Because of population growth and the aging of the US population, the absolute number of persons with cancer will likely continue to increase in the future.
Second, mortality from cancer, the second leading cause of death in the United States,2 has changed over time. Cancer-related deaths increased at an annual rate of 0.5% in the 15 years before 1990, whereas the 2002 to 2005 period showed an annual decline of 1.8%.1 Mortality is important from the standpoint of spending because a large proportion of healthcare costs are concentrated in the months before death. Trends in cancer death rates reflect trends in incidence rates, improvements in the effectiveness of early detection and treatment, and increases in longevity because of reductions in mortality from competing risks such as heart disease.
Third, improvements in cancer detection and treatment effectiveness (such as those seen in the treatment of breast cancer) are also partly responsible for changes in cancer spending.3 Fourth, trends in spending by payer type are affected by shifts in eligibility for and the benefit design of public insurance programs and employer-sponsored health insurance.
All of these factors are likely to have influenced the economic burden that cancer imposes on the US healthcare system. Although a few studies have quantified cancer costs for specific payers,4-7 there has not been a comprehensive analysis of how aggregate cancer costs have changed over time, and who now bears the burden of financing the bulk of these costs. This information is important for enhancing our understanding of the burden of cancer on specific payers, for budgeting purposes, for setting research and program priorities, and for use in cost-effectiveness analyses. The objective of our analysis was to use nationally representative data that span 2 decades to:
Present updated estimates of the prevalence and total cost of cancer for select payers (Medicare, Medicaid, other public, private, out of pocket) and how these have changed over the past 2 decades.
For each payer, present the distribution of payments over time by type of service (inpatient, prescription, outpatient/physician, and other) to assess whether there have been shifts in the settings in which cancer is treated.
Our results provide comprehensive estimates of the aggregate costs of cancer by payer and type of service within payer and how these costs have changed over time.
MATERIALS AND METHODS
We used data from the 2001 through 2005 Medical Expenditure Panel Survey (MEPS) and its predecessor, the National Medical Expenditure Survey (NMES), a 1-time survey conducted in 1987. Both surveys are nationally representative and capture self-reported data on medical conditions and related expenditures by type (eg, those related to prescription drugs) and source of payment (eg, Medicaid, private, self) for the US civilian noninstitutionalized population. Neither NMES nor MEPS include spending related to residents of long-term care facilities. Details of the surveys have been published previously.8 The costs captured by MEPS represent payments (not charges) from the payer to the provider. These are obtained through a combination of self-report (household component) and follow-back surveys with providers (medical provider component). NMES costs represent charges; we adjusted charges in NMES to be consistent with the payments in MEPS using methods developed by the Agency for Healthcare Research and Quality.9 Furthermore, both NMES and MEPS expenditures were inflated to 2007 US dollars using the annual gross domestic product price index.10
We identified medical conditions in MEPS and NMES based on similar questions in both surveys that ask participants to self-report conditions that required medical treatment at some point during the interview year. Medical conditions are not validated with provider reports from the medical provider component.11 Trained coders then classified self-reported medical conditions using International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9-CM) codes. Although NMES coded the self-reported conditions into more detailed 4-digit ICD-9 codes, we used only the first 3 digits for the medical conditions, to ensure consistency with MEPS. The MEPS and NMES samples included 135,714 individuals and 28,639 individuals, respectively; weighting variables were used to generate nationally representative estimates.
We used econometric methods to estimate cancer-attributable medical expenditures by payer (Medicare, Medicaid, private insurers, other public payers, and out of pocket) and type of service (inpatient, outpatient/office-based, prescription, and other). The “other” category included costs related to emergency room visits, home healthcare, vision aids, and other medical supplies. The econometric approach predicts medical expenditures at the individual level as a function of the presence or absence of cancer, other costly or prevalent medical conditions, and sociodemographic characteristics. We ran separate analyses for total medical expenditures by source of payer and type of service. In each model, the dependent variable was defined as annual expenditures for a given payer/service. The estimation sample for each payer population included all individuals who reported coverage by the payer or with a payment from that source.
As is typical with medical expenditure data, the samples included many individuals with zero expenditures during the year for some points of service and some individuals with extremely high expenditures. Therefore, we predicted annual expenditures for each survey using a 2-part model. The first part of the model used logistic regression to predict the probability of any medical expenditure for the payer/service. The second part of the model predicted expenditures only for those people who actually have positive expenditures. To select the appropriate cost estimation model for the second part of the 2-part model, we used the algorithm recommended by Manning and Mullahy for choosing among alternative nonlinear estimators.12 The algorithm indicated that we should use a generalized linear model with a gamma distribution and a log link. We multiplied the predicted probability from the first stage of the model with predicted (conditional) expenditures from the second stage to generate predicted annual medical expenditures for each individual in the database as a function of his or her disease profile and demographics.
Each part of the model included the same set of independent variables. The primary variable of interest was an indicator for cancer type (colorectal, lung, breast, prostate, skin, cervical, and other). Additional independent variables included the following high-cost or highly prevalent medical conditions: heart disease, stroke, congestive heart failure, hypertension, diabetes, injuries, dyslipidemia, human immunodeficiency virus/acquired immunodeficiency syndrome, pneumonia, asthma, chronic obstructive pulmonary disease, depression, other mental health/substance abuse, arthritis, back problems, skin disorders, renal failure, and pregnancy. All regressions also included as independent variables the following demographic characteristics: age, age squared, sex, race/ethnicity (white, black, Hispanic, Asian, other), education (missing degree, younger than 16 years, no degree, high school diploma, college degree, graduate degree, other degree such as an associate degree), family income (<100%, 100% to 199%, 200% to 399%, ≥400% of the poverty line), and, for the MEPS analyses, year indicators. All analyses were conducted using Stata software, version 10 (StataCorp, College Station, Tex).
We used the estimates from the econometric models to quantify cancer-attributable medical expenditures for each payer/service. Typically, disease-attributable expenditures are calculated by predicting expenditures using observed diseases and subtracting from that predicted expenditures, setting the disease of interest (ie, cancer) to zero and leaving all other covariates and diseases as they are in the data. However, we have shown in previous work that in nonlinear models, such as that used here, this approach will lead to double counting of expenditures for co-occurring diseases, regardless of whether 1 disease causes the other.13 Double counting of expenditures is a particular problem in cases where >1 condition is treated during a single office visit or hospitalization. We used a technique, termed complete classification and described in an earlier study, to ensure no double counting occurs.13 Specifically, using the parameters of the econometric model, we treated each disease and combination of diseases observed in the data as its own separate entity when calculating attributable costs. For example, cancer alone and cancer with heart disease would be treated as 2 different diseases in the attributable expenditure calculation described above. We then divided the total expenditures attributable to the combinations of diseases back to the constituent diseases using the parameters from the model to construct shares for each constituent disease within a combination (ie, a share of all cancer with heart disease costs that are attributable to cancer). The shares attribute a greater share of the joint expenditures to the disease with the larger coefficient in the main effect. The formula to construct the shares is given in Trogdon et al.13
We break down the change in total cancer costs in the following way. Let P05 (P87) and $05 ($87) be the number of prevalence cancer cases and the average cost per person with cancer in MEPS (NMES). Then, for each payer, the change in cancer costs is
The first term after the equal sign shows the change because of prevalence; it compares cancer costs using prevalent cancer cases from the 2 periods holding cost per person with cancer constant. The second term shows the change in cancer costs because of cost per person with cancer.
The prevalence of cancer cases increased among all payer populations between 1987 and the 2001 to 2005 period (Table 1). This, combined with an increase in the number of enrollees for all payers except “other” payers, led to substantial increases in the number of cancer cases, ranging from an 81.1% increase in the number of cases for patients covered by other public payers to a 250.8% increase in the number of cases for those covered by Medicaid.
|Payer||NMES||MEPSa||Increase in No. of Cases|
|Enrolled Population (in Millions)||Prevalence||No. of Cases (in Millions)||Enrolled Population (in Millions)||Prevalence||No. of Cases (in Millions)|
In 1987, the total medical cost of cancer (real-expressed in 2007 US dollars) was $24.7 billion (Table 2). Private payers financed the largest share of the total (42%), followed by Medicare (33%), patients paying out of pocket (17%), other public payers (7%), and Medicaid (1%) (Fig. 1). Between 1987 and the 2001 to 2005 period, the total (real) medical cost of cancer increased by 98%, to $48.1 billion. However, as a share of overall medical expenditures, cancer costs remained relatively constant (4.8% of $513 billion in total medical expenditures in 1987 and 4.9% of $979 billion during 2001 to 2005). Thus, the increase in cancer costs matched the overall increase in medical expenditures during the last 20 years.
|Payer||Type of Service||NMES||MEPSa|
|Cancer Costs (Millions in 2007 US$)||Cancer Costs (% of Total Expenditures)||% of Cancer Costs Within Payer||Cancer Costs (Millions in 2007 US$)||Cancer Costs (% of Total Expenditures)||% of Cancer Costs Within Payer|
However, the increase in cancer expenditures varied substantially across payers. In real terms, Medicaid costs increased by 488% ($1.3 billion), costs to private insurers increased by 137% ($13.8 billion), and Medicare costs increased by 99% ($8.1 billion). Costs paid by other public programs increased by a much smaller amount: 25% ($459 million). Costs paid out of pocket, which include copayments and deductibles plus costs for the uninsured, fell by 7% ($288 million). As with changes across payers, the increases in cancer costs within payer also tended to match the increase in overall payer expenditures; cancer costs as a share of total payer expenditures were generally stable (eg, 5.3% of private insurance expenditures were attributable to cancer in 1987 and 5.7% in the 2001 to 2005 period).
As a result of the changes described in the preceding paragraph, the distribution of cancer-related costs across payers has shifted (Fig. 2). The share of cancer costs paid by private insurance increased from 42% to 50%, whereas the share of out-of-pocket costs fell from 17% to 8%; other payers' share fell from 7% to 5%. Even with an aging population, the share of cancer costs paid for by Medicare remained fairly constant over this time period (33% vs 34%). The share of cancer costs paid by Medicaid remained low but tripled from 1% to 3%.
As with other medical services, another major trend that occurred over this time period was a movement away from cancer treatment in the inpatient setting. Across payers, the share of cancer costs that resulted from inpatient admissions fell from 64.4% of total cancer costs in 1987 to 27.5% during the 2001 to 2005 period. This shift occurred within every payer, although to differing degrees. In real terms, total inpatient expenditures dropped from $15.9 billion to $13.2 billion over this time period.
As shown in Table 3, the reduction in the percentage of expenditures resulting from inpatient cancer treatment can be explained in part by a decrease in the percentage of cancer patients admitted to hospital for care and decreased average length of hospital stay. The probability of receiving inpatient cancer treatment in the past year among those with a positive cancer diagnosis decreased for every payer between 1987 and the 2001 to 2005 period, ranging from a 34% decrease for Medicaid (from 40.7% to 26.9% of cancer patients) to a 56% decrease for out-of-pocket payers, including the uninsured (from 22.7% to 9.9%). However, among patients who were admitted, average length of stay was relatively stable.
|Payer||Cancer Patients Admitted to Hospital for Cancer||Average Length of Stay Given Admission, d|
For every type of payer, the decrease in cancer-related inpatient expenditures has been accompanied by an increase in cancer-attributable outpatient expenditures (Table 2). Some payers also experienced an increase in cancer-attributable prescription drug expenditures. In both time periods, prescription drug expenditures administered outside of the inpatient setting accounted for a low percentage of all cancer-attributable costs paid by Medicare and other public payers. Cancer-attributable spending for prescription drugs increased for patients with private insurance (from 1.4% to 7.0%) and for out-of-pocket payers (from 5.2% to 27.7%) across the 2 time periods. Across all payers, prescriptions accounted for 6.1% of all cancer-attributable medical expenditures in the 2001 to 2005 period, up from 1.8% in 1987.
For Medicaid and private insurance, 86% and 84% of the overall increase in costs was because of the change in prevalence (see Eq. 1). For Medicare and other public insurance, the change because of cost per person was actually negative (ie, if the number of cases had remained the same, the costs of cancer would have decreased), indicating that the increase in prevalence more than made up for a decrease in costs per person. Out of pocket was the only insurance group in which the change in cancer costs (a net decrease between NMES and MEPS) was mainly driven by changes in costs per person; costs per person decreased by enough to offset an increase in the number of prevalence cases.
The results of this study reveal 3 important trends in the costs of cancer over the past 20 years: 1) the total medical costs of cancer have nearly doubled, although this increase is not dissimilar to overall trends in aggregate health spending; 2) as with general medical services, cancer-related treatment costs have shifted away from the inpatient setting and toward the outpatient setting; and 3) the share of these costs paid for by private insurance and Medicaid has increased.
Cancer-specific medical expenditures have doubled in the past 20 years. Comparing the results from Tables 1 and 2, the number of cancer cases has increased proportionately more than medical expenditures, indicating that the increase in cancer costs has been driven mainly by the increase in the number of cases and not by the cost per treated cancer case. The dramatic increase in the number of cancer cases is because of new cases diagnosed among the aging population base, as well as increasing cancer prevalence. Despite a downward trend in overall cancer incidence, cancer prevalence has increased because of more cancers occurring in an aging population and, for some cancers, improved treatments leading to improved survival rates.1, 3
The upward pressure on cancer costs caused by the increase in the number of people living with cancer has been offset somewhat by a slight decrease in the average medical expenditures per patient. This is the result of the shift away from inpatient treatment toward the outpatient setting, as seen in Table 2 and in previous studies.14, 15 For example, many cancer patients now receive chemotherapy and radiation therapy in the outpatient clinic setting.16 Advances in medicine and technology, changing demographics, and legislation have contributed to this shift.15 Other factors driving the change are cost-containment strategies implemented by payers, limited inpatient support resources, improved outpatient supportive care, and patient preference for shorter hospital stays.14 All told, these shifts did not outweigh the increases in costs caused by increased cancer prevalence over the past 20 years. The combination of the upward trend in prevalent cancer cases with changes in treatment settings resulted in cancer costs increasing at the same rate as overall medical expenditures. Over the past 20 years, nearly 5% of all medical spending in the United States has gone to the care of cancer patients (Table 2).
Whereas the share of medical expenditures going to cancer has remained constant, the burden to payers has changed. Private insurance, which was already paying for the largest share of cancer costs in 1987, has seen its share of the cancer costs burden increase to half of all cancer-attributable medical expenditures. Despite a decrease in the price of private insurance,17 population increases have led to an increase in the total number of the privately insured (Table 1). As cancer screening rates in the pre-Medicare population continue to increase, cancer-attributable costs to private payers will likely continue to rise. Because early detection has shifted an increased proportion of the burden to private payers, we can expect that private payers will take a stronger interest in the quality and costs of cancer care.
Because of the relatively young population served by Medicaid, its share of cancer costs is small, but Medicaid experienced the largest relative increase. This is because of the dramatic increase in the number of cancer patients covered by Medicaid and an increase in expenditures per treated patient (results not shown). The latter effect is likely the result of Medicaid paying for cancers detected at later stages and/or cancer patients with comorbidities who need inpatient care.18
Our analysis extends previous estimates of the medical cost of cancer in 2 ways. First, we present estimates for all payers and use more recent data than were available in previous studies.6, 7 One of those studies reported that cancer spending accounted for 5.4% of all medical expenditures among the nonelderly in the 1996 to 1999 MEPS, which concurs with our finding of relatively stable cancer spending as a percentage of overall medical spending.7 Second, we presented trends in spending by payer and type of service over the last 20 years. Similar figures have been reported from the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database for 1991 through 20024; our study reports cancer costs for other payers in addition to Medicare.5
Although the NMES and MEPS data allowed us to assess the trends in the cost of cancer over the past 20 years, these data sources also have limitations. First, they capture only a fraction of total national health expenditures. For example, both exclude spending related to residents of long-term care facilities.19 In the authors' analysis of the 2004 National Nursing Home Survey, cancer patients were not over-represented in nursing homes. However, estimates reported here do not include nursing home expenditures attributable to cancer. In general, differences in the relative proportion of different cancer types over time as well as changes in medical practice patterns will influence overall treatment patterns and related costs. Analysis of more specific cancer sites (eg, breast or colorectal cancer) was not possible because of small sample sizes. In addition, medical conditions in NMES and MEPS are self-reported and not verified by chart review. Johnson and Sanchez11 compared self-reported conditions by households to the reports of their medical providers in NMES. Defining agreement as a match between any condition listed as a reason for a medical event by the household and any disease listed by the medical provider for the same event, the rate of agreement for neoplasms was 43% at the 3-digit ICD-9 level and 55% at higher levels of groupings. The limitations of the self-reported data must be balanced against the benefits provided by NMES/MEPS relative to SEER-Medicare, namely that nationally representative cost-of-illness estimates can be estimated for the non-Medicare population and payers. Although self-reporting can be problematic for some types of conditions, in general the literature supports the validity of this approach.20, 21 Cancer patients with advanced disease may be unlikely to participate in either of these surveys, further contributing to underestimation of cancer costs. Finally, the true burden of cancer includes nonmedical costs for travel and child care, costs incurred by caretakers, and nontangible costs associated with psychological pain and stress, as well as lost productivity.22, 23 Although medical costs are often the sole focus of economic research and debate, a full accounting of the cost and benefits of policy changes affecting cancer patients should take such nonmedical costs into consideration.
The future of cancer detection and treatment is certain to involve changes that affect not only cancer-related medical expenditures but also which payers are responsible for those expenditures. Out-of-pocket expenditures associated with cancer treatment are likely to increase because of increases in the cost of cancer-treatment procedures and drugs24 and the trend toward consumer-driven healthcare options, such as Health Savings Accounts and High-Deductible Health Plans, which require greater levels of consumer cost-sharing.25 Conversely, proposed healthcare reform measures would offer health insurance coverage to those who are currently uninsured, which would shift cancer expenditures further to private insurance or to public payers, depending on how the coverage is extended. Unless significant reductions in reimbursement rates are negotiated, a public option will likely increase the costs to public health insurance plans as the number of persons covered by the government increases. Policy changes to public programs are also likely to shift the burden of cancer spending. For example, the National Breast and Cervical Cancer Early Detection Program and the Breast and Cervical Cancer Prevention and Treatment Act have been expanded to provide an additional route for women with breast and cervical cancer to gain Medicaid eligibility.26 The information provided in this study enhances our understanding of the burden of cancer on specific payers and how this burden may change as a result of healthcare reform measures or other changes to healthcare financing and delivery. Further research will be needed to determine the impact of these and other changes on costs and quality of cancer care in the United States.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.
- 2RiesLAG, MelbertD, KrapchoM, et al, eds. SEER Cancer Statistics Review, 1975-2005. Bethesda, MD: National Cancer Institute; 2008. Available at: http://seer.cancer.gov/csr/1975_2005/ Accessed September 25, 2008.
- 6Health insurance and spending among cancer patients. Health Aff (Millwood). 2003; W3: 189-198., .
- 10United States Department of Commerce, Bureau of Economic Analysis. Available at: http://www.bea.gov/national/index.htm Accessed January 6, 2008.
- 11Household and medical provider reports on medical conditions: National Medical Expenditure Survey, 1987. J Econ Soc Meas. 1993; 19: 199-223., .
- 17Income, poverty, and health insurance coverage in the United States: 2005. US Census Bureau, Current Population Reports, P60-231. Washington, DC: US Government Printing Office; 2006., , .
- 22Cost analysis of screening for, diagnosing, and staging prostate cancer based on a systematic review of published studies. Prev Chronic Dis. 2007; 4: A110., , .
- 25AHIP Center for Policy and Research. 2008 HSA Census. Available at: http://www.ahipresearch.org/pdfs/2008_HSA_Census.pdf Accessed September 15, 2008.
- 26Centers for Disease Control and Prevention. Breast and Cervical Cancer Prevention and Treatment Act of 2000. Available at: http://www.cdc.gov/cancer/nbccedp/legislation/law106-354.htm Accessed February 24, 2009.