Due to scientific advances in oncology, increased use of more expensive cancer treatments, and an aging population, the cost of cancer care is expected to increase enormously. In 2010, the national cost of cancer care was estimated to be $124.6 billion USD in the SEER-Medicare population, of which female breast cancer (BC) constituted the highest proportion at 13% ($16.5 billion). Further, it is projected to increase by 32% in 2020, presenting a significant burden to Medicare.1 Of the $16.5 billion, the initial phase of care (12 months) following BC diagnosis represented 37% of the cost, the continuing phase of care 41%, and the last year of life accounted for 22%. BC costs are substantially higher for the initial phase of care due to surgery, radiotherapy (RT), and adjuvant therapy.2–6 Regardless of these extensive costs to Medicare, there is insufficient up-to-date information on healthcare utilization and costs incurred by specific services, and factors significantly contributing to these costs in women aged ≥65 years. Because older women have higher BC incidence than their younger counterparts,7 it is important to determine the economic burden of BC in planning for future healthcare resource allocation.
A descriptive review about the costs of cancer care in the United States reported that studies varied widely regarding settings, methodologies, cost perspectives, populations, measurements of costs, types of services included, time horizons, and data sources.8 Another systematic review included outdated studies, which may not reflect changes in the patterns of care due to technological advances and innovations in BC treatment.3 Furthermore, other studies used data from fewer cancer registries, and not for all BC stages.1,2,6,9,10 Additionally, costs according to types of specific services have not been reported.6,10 One study that determined costs during the initial phase of care for Medicare beneficiaries with BC in Virginia reported comorbidity, hospital stay, and type of treatment as the significant cost contributors.11 Another study that used SEER-Medicare data reported higher healthcare utilization, although the study findings had limited generalizability due to inclusion of only 4 SEER geographic areas.9 In fact, to date, no single comprehensive study focusing on both healthcare utilization and costs during the initial phase of care in older women with BC from all SEER areas has been published with the results stratified by specific services.
The purpose of this study was to determine the treatment approaches and healthcare utilization and costs during the initial phase of BC care among female Medicare fee-for-service beneficiaries using the SEER-Medicare data, and to determine factors associated with costs using a multivariate framework.
Study Design and Data Source
A retrospective observational cohort study was conducted from the Medicare perspective using data from the SEER program, which routinely gathers data on socioeconomic and clinical factors about cancer cases from 17 population-based tumor registries that in turn collect data from hospitals, outpatient clinics, laboratories, private practitioners, hospices, autopsy reports, and death certificates; it represents 26% of the US population.12 Approximately 94% of SEER cases were matched to their Medicare claims from the time of their eligibility until death based on name, social security number, sex, and date of birth.13 This study used information from the Medicare claims files, which were linked to the SEER cancer cases.14 More details of the SEER-Medicare data set are described elsewhere.12 For this study, the Area Resource File was linked to the SEER-Medicare data set using the state and county Federal Information Processing Standards code for each beneficiary in order to obtain the county level information on income, education, healthcare providers, and hospitals offering oncology services.15
The study cohort comprised women aged ≥66 years with first primary incident BC (ICD-9-CM codes 174.xx, 233.0x, 238.3x, 239.3x) during 2003–2009. Among a total of 195,067 women identified, the following women were excluded: 11,031 who died within 12 months after BC diagnosis; 71,254 aged <66 years; 5,524 with any previous cancer diagnosis; 5 who were diagnosed with BC during death or autopsy; 2,072 with unknown or missing BC stage; 6,299 not continuously enrolled in Medicare Parts A and B in the 12 months before and after diagnosis; 29,195 who were members of a health maintenance organization at any time during the study period; and 380 who had zero Medicare costs. Therefore, 69,307 women met inclusion criteria.
Initial Phase of Care: Based on the consensus from a group of clinical oncologists, the initial phase of care was defined as 12 months after BC diagnosis; initial courses of treatment including adjuvant therapy were usually completed within 1 year of cancer diagnosis; this definition is used for comparison purposes with previous literature.1,2,4–6,11,16
Healthcare Utilization: Healthcare utilization included inpatient visits, hospitalization days, outpatient visits, and emergency department (ED) visits derived from Medicare claims files. Additionally, the percentage of women who received surgery, RT, or chemotherapy was calculated. The average number of RT and chemotherapy visits were determined.
Cost Estimates: All of the Medicare files, including inpatient (Medicare Provider Analysis and Review [MedPAR]), hospital outpatient, carrier/National Claims History (NCH), hospice, home health agency (HHA), and durable medical equipment (DME) files, were used to obtain direct Medicare costs. Costs were defined as the amount reimbursed by Medicare. Reimbursement, used as a proxy for Medicare costs in previous cost studies,4,5,10 are actual payments derived from reimbursement formulas reflecting the average resource utilization for each healthcare service.17 Total costs, costs within the categories of care, such as inpatient services obtained from MedPAR file, outpatient services obtained from outpatient file, physician services obtained from NCH file, and other services (DME, hospice care, and HHA), were calculated to identify which Medicare component contributed substantially to total costs. All costs were adjusted using the previously used method6 and were reported in 2012 $USD to account for variation over time.18
Independent Variables: To identify the factors associated with costs, the Andersen behavioral model of healthcare services utilization was used.19,20 According to this model, use of healthcare services is a function of people's predisposition to use services, factors that enable or impede use, need for care, and factors associated with healthcare use and external environmental healthcare-related factors. Predisposing factors included age at BC diagnosis and race; whereas enabling factors included marital status, census tract median household income, and census tract percentage of people aged ≥25 years with at least 4 years of college education. Need-related factors related to specific cancer care included stage at diagnosis,21 grade of tumor, estrogen receptor (ER) status, comorbidity score,22,23 and mental health conditions (ie, depression and/or anxiety) derived from co-occurring chronic conditions within 12 months before BC diagnosis. Factors associated with healthcare use were the number of primary care physician (PCP) visits in the year prior to BC diagnosis,24 type of initial treatment in the year of BC diagnosis (ie, definitive surgery only, nonsurgical treatment [chemotherapy, RT, or both], definitive surgery plus non-surgical treatment, no treatment), and inpatient use. External environmental healthcare-related factors included location of residence, SEER region, and number of hospitals offering oncology-related services in the area of residence.
Descriptive statistics were used to describe the characteristics of the study cohort. Average healthcare utilization was calculated for each service, as well as for RT and chemotherapy. Percentages were determined of women who had surgery, chemotherapy, or RT, and those who had at least one inpatient, outpatient, or ED visit. To understand the trends in costs over time, average total costs and costs according to specific services during the initial phase of BC care were calculated and reported for each calendar year of diagnosis.
The Breusch-Pagan/Cook-Weisberg test and simplified White test were performed to report the heteroscedasticity and kurtosis of log-scale residuals in the cost data.25 Due to distributional problems of cost data, Park tests26 were conducted to determine the most appropriate regression model for data based on which generalized linear model (GLM) with log link function and gamma distribution was conducted to model costs and identify its predictors in a multivariate framework. From the regression, the regression estimates (betas) were exponentiated to yield average costs. The findings with P values ≤0.05 levels are discussed. All analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC) and Stata version 13 (StataCorp LP, College Station, TX).
Table 1 describes the study cohort along with the average costs during the initial phase of care within each group. Most patients were white (88%), resided in metropolitan areas (84%), were diagnosed with local-stage BC (58%), had ER-positive status (72%), and had no mental condition (94%) or comorbidity (57%). Average costs were higher among women who were black; resided in metropolitan areas; were diagnosed with distant, poorly differentiated, and
Description of Study Cohort (N=69,307)
The upper part of Table 2 describes healthcare utilization wherein surgical intervention was reported for 96% of women. The mean number of chemotherapy visits was 2.5 (SD, 6.9), and 21% of the study cohort used this treatment at least once during the initial phase of care. The mean number of RT visits was 6.2 (SD, 8.0), and 54% used RT. Almost all patients (99.9%) reported at least one outpatient visit, with an average of 40.1 visits (SD, 20.3); 45% of the cohort reported at least one inpatient visit, with an average of 0.7 visits (SD, 1.1) and an average of 4.7 hospitalization days (SD, 14.8). Almost a quarter of the cohort (24%) reported ≥1 ED visit with an average of 0.4 visits (SD, 1.0).
Average Costs and Associated Trends
The lower part of Table 2 describes average costs and costs by types of specific services. Average costs were $28,075 (SD, $23,978). Physician services contributed to the highest proportion (47.5%) of the costs ($13,344; SD, $13,773), whereas inpatient and outpatient services contributed to 21.5% and 26.6% of the costs, respectively, with corresponding amounts of $6,046 (SD, $13,905) and $7,456 (SD, $8,593), respectively.
Figure 1 describes trends in the average total costs and costs by types of specific services. From 2003 through 2009, there was an increase in average total costs (from $26,361 to $29,522, respectively), average costs due to outpatient services (from $6,447 to $8,730, respectively), and average costs due to physician services (from $12,032 to $13,946, respectively), whereas the average costs due to inpatient services declined (from $6,620 to $5,602, respectively).
Costs were higher in the study cohort with advanced-stage versus early-stage BC: $18,354 among those with in situ BC; $25,085 for those with local-stage BC; $40,014 for those with regional-stage BC; and $52,182 for those with distant-stage BC (data not shown). Moreover, costs for specific services (inpatient and outpatient services, physician services, and other services) increased with advancing stage at diagnosis.
Factors Associated With Average Costs
Table 3 summarizes results from the GLM regression on costs. Women who were diagnosed at advanced stages of BC, did not have a well-differentiated tumor grade, had ER-negative tumor status, had treatment in the form of surgery or adjuvant therapy, reported inpatient use, had higher PCP visits, and had any comorbidity and/or mental condition had significantly higher costs during the initial phase. Women who were black, resided in areas with at least 4 hospitals offering oncology services, lived in the Northeast SEER region, and lived in areas where the household income was ≥$50,000 had higher costs as well. In addition, women who were aged ≥70 years, were of “other” race, were married/partnered, and resided in nonmetropolitan areas and in the South and North Central SEER regions had significantly lower costs during the initial phase of care.
As costs of BC treatment constitute a substantial portion of overall national cancer costs and are projected to increase tremendously in the coming years,1 estimating average costs and costs by types of specific services is critical for Medicare to identify cost drivers. This study is the first of its kind to use more recent SEER-Medicare data from 2003–2009 BC cases to estimate healthcare utilization, average costs, costs by types of specific services, and the significant
Healthcare Utilization and Costs by Type of Specific Service
The percentage of women who underwent surgery during the initial phase of care was significantly higher than that reported in a previous study (96% vs 91%), whereas the percentages who underwent chemotherapy, RT, and hospitalization remained nearly consistent with the previously published data (21%, 54%, and 45%, respectively).10 The average cost during the initial phase of care in older women with incident BC was $28,075, which is significantly higher than the costs reported in previous studies.6,9,11 Before and during our study period, several new BC drugs were approved by the FDA (eg, letrozole in 2001, fulvestrant in 2002, gemcitabine in 2004, nab-paclitaxel in 2005, lapatinib and ixabepilone in 2007),27 which may be available at branded price during the study period, thereby suggesting increased Medicare expenditures.28 In addition, use of more expensive radiation technology, such as brachytherapy and intensity-modulated RT,29 may also cause higher Medicare expenditures.
In contrast to the literature,4,10 our study found that physician services followed by outpatient services were the major contributors to the average total costs indicating an increased breast-conserving therapies30,31 and new adjuvant therapies that require more frequent monitoring as reflected by higher adjusted costs associated with surgery and/or adjuvant therapy in Table 3. These findings likely represent several changes in treatment patterns, including increased
Factors Associated With Costs Using Generalized Linear Model Regression
Among predisposing factors, age and race were significant predictors of cost. Increasing age at diagnosis resulted in decreasing costs, a similar finding as reported in published literature,5 perhaps because older women are less likely to receive aggressive treatment after BC diagnosis, resulting in lower costs.16,36,37 Additionally, consistent with a previous study,5 higher costs were reported for African American women, perhaps because they have been shown to receive less preventive and diagnostic medical care38 and hence are more likely to use inpatient services, resulting in higher costs.39 Warren et al5 suggested that African American women may receive different treatment compared with white women, and that certain unmeasured factors (eg, body mass index, health status) may affect differences in care, thereby affecting costs. Among enabling factors, living in areas with higher household income was independently associated with higher costs, a similar finding as reported in a previous study.11
With regard to need-related factors, advanced disease stage, moderately and poorly differentiated tumor grade, ER-negative tumor status, and presence of chronic physical and mental conditions were significantly associated with higher costs. These findings are consistent with those of previous studies reporting significant increases in costs associated with advancing disease stages4,5,11,16 and the presence of comorbidities,11 suggesting the importance of co-management of physical and mental chronic conditions40,41 and improving access to preventive cancer screening services for older women. Emerging healthcare delivery models, such as “medical homes,” that emphasize care coordination for Medicare beneficiaries with multiple chronic conditions could be a step in the right direction to curb increasing costs associated with BC care. Among factors associated with healthcare use, women with BC who had surgery and adjuvant therapy and those who had nonsurgical treatment had higher costs than those who did not have any BC treatment, as reported by Penberthy et al.11 Furthermore, inpatient use and PCP visits were also significantly associated with higher costs. Older women with BC who resided in metropolitan areas and had access to a higher number of hospitals providing oncology services had higher costs in the initial phase of care, perhaps because of increased access to Medicare insurance.
Several limitations of this study are worth noting. Healthcare services not reimbursable by Medicare may not be captured, and healthcare utilization and costs of adjuvant therapy did not capture prescription drugs, potentially resulting in underestimation of cost. Data on variables such as health status, severity of comorbidities, and patient preferences, which may impact cancer prognosis and/or selection of BC treatment, were not available. Moreover, the study findings are generalizable to older patients with BC covered by Medicare.
This population-based study suggests that BC represents a substantial economic burden during the initial phase of care. Physician and outpatient services accounted for the highest proportion of costs. Predisposing factors, need-related factors, healthcare use, and external environmental healthcare-related factors significantly predicted costs.
The authors have disclosed that they have no financial interests, arrangements, affiliations, or commercial interests with the manufacturers of any products discussed in this article or their competitors.
This study was part of Dr. Vyas' doctoral dissertation at West Virginia University and was supported by an AHRQ grant (R24HS018622-03). Some additional salary support was received by Drs. Madhavan and Sambamoorthi from a National Institute of General Medicine Sciences (NIGMS) IDeA Clinical and Translational Research grant (U54GM104942) obtained by West Virginia University. The content is solely the responsibility of the authors and does not necessarily represent the official views of AHRQ and NIGMS.
Partial results of this analyses were presented at the International Society for Pharmacoeconomics and Outcomes Research 20th Annual International Meeting; Philadelphia, PA; May 16–20, 2015.
BrownMLRileyGFSchusslerNEtzioniR. Estimating health care costs related to cancer treatment from SEER-Medicare data. Med Care2002;40(8 Suppl):IV-104–117.
CampbellJDRamseySD. The costs of treating breast cancer in the US: a synthesis of published evidence. Pharmacoeconomics2009;27:199–209.
RileyGFPotoskyALLubitzJDKesslerLG. Medicare payments from diagnosis to death for elderly cancer patients by stage at diagnosis. Med Care1995;33:828–841.
WarrenJLBrownMLFayMP. Costs of treatment for elderly women with early-stage breast cancer in fee-for-service settings. J Clin Oncol2002;20:307–316.
HowladerNNooneAMKrapchoM eds. SEER Cancer Statistics Review 1975-2013National Cancer Institute. Bethesda, MD based on November 2011 SEER data submission April2012. Available at: http://seer.cancer.gov/csr/1975_2013/. Accessed January 21 2017.
HanchateADClough-GorrKMAshAS. Longitudinal patterns in survival, comorbidity, healthcare utilization and quality of care among older women following breast cancer diagnosis. J Gen Intern Med2010;25:1045–1050.
PenberthyLRetchinSMMcDonaldMK. Predictors of Medicare costs in elderly beneficiaries with breast, colorectal, lung, or prostate cancer. Health Care Manag Sci1999;2:149–160.
WarrenJLKlabundeCNSchragD. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care2002;40(8 Suppl):IV-3–18.
EngelsEAPfeifferRMRickerW. Use of Surveillance, Epidemiology, and End Results-Medicare data to conduct case-control studies of cancer among the US elderly. Am J Epidemiol2011;174:860–870.
National Cancer Institute Division of Cancer Control & Population Sciences. SEER-Medicare: About the Data Files. Available at: http://healthcaredelivery.cancer.gov/seermedicare/aboutdata/. Accessed January 21 2017.
U.S. Health Resources and Services Administration. Bureau of Health Professions. Area Resource File 2009–2010 Release. Rockville, MD: US Department of Health and Human Services; Fairfax, VA: Quality Resource Systems, Inc; 2010. Codebook: HE-001(2010).
TaplinSHBarlowWUrbanN. Stage, age, comorbidity, and direct costs of colon, prostate, and breast cancer care. J Natl Cancer Inst1995;87:417–426.
United States Department of LaborBureau of Labor Statistics. Consumer price index medical care services 1982-84. Available at: https://www.bls.gov/cpi/. Accessed January 21 2017.
AndersenRNewmanJF. Societal and individual determinants of medical care utilization in the United States. Milbank Mem Fund Q Health Soc1973;51:95–124.
KlabundeCNPotoskyALLeglerJMWarrenJL. Development of a comorbidity index using physician claims data. J Clin Epidemiol2000;53:1258–1267.
National Cancer Institute Division of Cancer Control & Population Sciences. SEER-Medicare: Calculation of Comorbidity Weights. Available at: https://healthcaredelivery.cancer.gov/seermedicare/considerations/calculation.html. Accessed January 21 2017.
YuXMcBeanAMVirnigBA. Physician visits, patient comorbidities, and mammography use among elderly colorectal cancer survivors. J Cancer Surviv2007;1:275–282.
FDA Approved Drugs for Oncology: Drugs Approved in 2017. Available at: https://www.centerwatch.com/drug-information/fda-approved-drugs/therapeutic-area/12/oncology. Accessed April 28 2017.
HowardDHChernewMEAbdelgawadT. New anticancer drugs associated with large increases in costs and life expectancy. Health Affairs2016;35:1581–1587.
RobertsKBSoulosPRHerrinJ. The adoption of new adjuvant radiation therapy modalities among Medicare beneficiaries with breast cancer: clinical correlates and cost implications. Int J Radiat Oncol Biol Phy2013;85:1186–1192.
YangRLNewmanASLinIC. Trends in immediate breast reconstruction across insurance groups after enactment of breast cancer legislation. Cancer2013;119:2462–2468.
KennedyTStewartAKBilimoriaKY. Treatment trends and factors associated with survival in T1aN0 and T1bN0 breast cancer patients. Ann Surg Oncol2007;14:2918–2927.
ShihYCEltingLSHalpernMT. Factors associated with immunotherapy use among newly diagnosed cancer patients. Med Care2009;47:948–958.
HalpernMTYabroffKR. Prevalence of outpatient cancer treatment in the United States: estimates from the Medical Expenditures Panel Survey (MEPS). Cancer Invest2008;26:647–651.
SillimanRATroyanSLGuadagnoliE. The impact of age, marital status, and physician-patient interactions on the care of older women with breast carcinoma. Cancer1997;80:1326–1334.
EscarceJJEpsteinKRColbyDC. Racial differences in the elderly's use of medical procedures and diagnostic tests. Am J Public Health1993;83:948–954.
GornickMEEggersPWReillyTW. Effects of race and income on mortality and use of services among Medicare beneficiaries. N Engl J Med1996;335:791–799.
SubramanianSTangkaFKSabatinoSA. Impact of chronic conditions on the cost of cancer care for Medicaid beneficiaries. Medicare Medicaid Res Rev2013;2:pii: mmrr.002.04.a07.