Economic Burden of Melanoma in the Elderly Population: Population-Based Analysis of the Surveillance, Epidemiology, and End Results (SEER)–Medicare Data

Objective  To assess health care resources consumed by melanoma in the population 65 years or older, a group with comparatively poor outcomes.

Design  Database analysis.

Setting  The Surveillance, Epidemiology, and End Results (SEER)–Medicare-linked population-based database for fiscal years 1991 through 1996.

Participants  A total of 1858 subjects with pathological confirmation of melanoma.

Main Outcome Measures  Resource consumption was examined by stage and treatment phase. Outcomes were measured in monthly charges obtained from the data set and costs were estimated by application of cost to charge ratios. Annual resource consumption by melanoma in patients 65 years or older in the United States was also estimated by incorporation of published SEER cancer statistics.

Results  Average monthly, per-patient melanoma charges were $2194 during the initial 4 months of treatment; they dropped by more than half to $902 during the interim phase, which varied in length depending on survival. Monthly charges increased to $3933 during the terminal 6 months of treatment. The estimated annual charge of treating melanoma in the population 65 years or older was $390 million. By using cost to charge ratios, we found the annual cost of melanoma to be up to $249 million and the per-patient lifetime costs to be $28 210 from the time of diagnosis to the time of death.

Conclusions  Melanoma care presents a significant economic burden in the elderly population, particularly in late-stage disease. If effective, prevention and early detection efforts may reduce the economic burden.

Melanoma poses a significant health threat as the sixth most commonly diagnosed cancer in the United States.¹ Melanoma will affect approximately 1 in 49 men and 1 in 73 women¹ compared with the lifetime risk of 1 in 1500 persons in the 1930s.^2,3 Additional health care resources are being used, and more information on consumption will better inform policymakers.

The elderly population is of particular interest because of the disproportionate increase in deaths in patients 65 years or older.⁴ Older white men have the highest mortality from melanoma, and, as such, they may benefit from targeted detection efforts that might improve their outcomes.⁵ Assessing health care consumed in the elderly will serve as an important measure of economic burden that can be used to compare the burden of melanoma with that of other malignant neoplasms. In addition, knowledge of resources consumed will provide a baseline framework for assessing the impact of future interventions.

We sought to assess health care resources consumed by patients 65 years or older with melanoma. Our sample accounts for 12% of patients 65 years or older with melanoma in the United States. There were 2 primary goals in examining this group: (1) to determine the amount expended by stage and treatment phase and (2) to expand our data to the entire US population 65 years or older by applying cost data from our sample to national incidence and death rate statistics. This second goal was to estimate the annual cost of melanoma in all US patients 65 years or older. The estimate of national annual resources consumed allows for comparison with cost data from other malignant neoplasms.

We gathered population-based claims data from the Surveillance, Epidemiology, and End Results (SEER)–Medicare database. Institutional review board approval was waived because the SEER-Medicare data lack personal identifiers. Outcomes were measured in charges and costs. Results were categorized by phase (or time interval) after diagnosis, according to the method described by Brown et al.⁶

Charges were melanoma-specific and were derived directly from the database as an estimate of health care resources consumed. Each charge reflects that a designated service was provided. Although reimbursement amounts reflect what was paid for and, thus, “costs,” they may not capture services that were provided but not reimbursed or only partially covered. Furthermore, they may depend on previous services used and copayment amounts.

We applied cost to charge ratios to yield cost estimates. The derivation of costs was for comparison of our findings with those from previous studies examining costs in the general melanoma population (including all age groups) and in other malignant neoplasms. Our phase costs were combined with national estimates on incidence and mortality to approximate the annual direct cost of melanoma in the US population 65 years or older.

Data were gathered from the SEER-Medicare data set (fiscal years 1991-1996). The SEER files included all incident cancer cases in 5 states and 7 metropolitan areas in the United States, accounting for 15% to 20% of the population. The SEER-Medicare database is the only population-based data set in the United States that contains claims records for covered services from the time of eligibility until death. For each linkage (years 1991, 1995, and 1999), 93% of persons 65 years or older in SEER were also found in the Medicare files.⁷

Our sample was derived from 2020 patients with pathologically confirmed melanoma (International Classification of Diseases, Ninth Revision [ICD-9] code 172.x) used in a previous study from our group.⁸ For the current investigation, we excluded an additional set of patients whose tumors could not be categorized by stage, leaving 1858 patients. It is important to note that this previous study⁸ was concerned with provider type, so 17% of the SEER melanoma cases were excluded if the provider making the diagnosis was not a dermatologist or other practitioner (such as a primary care physician) who would be likely to perform a skin examination.

The SEER database provides tumor information according to the American Joint Committee on Cancer (AJCC) staging system, extent of disease, and historic stage definitions. We converted available information to AJCC staging criteria for uniformity. However, we were unable to convert the data to 2002 AJCC stages⁹ because of a lack of information on histologic ulceration for many patients. Accordingly, 1997 AJCC stages¹⁰ were used. A discussion of this method is available elsewhere.⁸

Charges were derived from 5 Medicare claims files: (1) Medicare Provider Analysis and Review (hospital claims); (2) Physician/Supplier Part B (physician claims); (3) Outpatient Claims (claims from institutional outpatient providers); and (4) Home Health Agency and (5) Hospice (claims for end-of-life care). We queried all files for melanoma ICD-9 diagnosis codes (172.x). We queried other lesions to take into account surveillance, because patients with melanoma are closely followed up and can have multiple biopsies of lesions that may turn out to be benign or involve a nonmelanoma result. For instance, we included carcinoma in situ of the skin (ICD-9 232.x), lentigo maligna (ICD-9 232.9 and 238.2), dysplastic nevus (ICD-9 238.9), and benign skin lesions (ICD-9 216.x, 702.11, and 702.19).

The Medicare Provider Analysis and Review, Physician/Supplier Part B, and Outpatient Claims were queried for Current Procedural Terminology codes (Table 1). Procedures were included only when melanoma or the alternative diagnosis codes, as listed previously, were designated. Our list of included codes was purposely broad. Although many of the procedures are not routinely ordered for patients with melanoma, they have been included in previous cost analyses.¹¹ For example, ultrasonography may be useful for evaluating lymph nodes in locations difficult to palpate, but this technique is not widely used in the United States.¹² Our overall approach was conservative in that we included only patients with pathologically confirmed melanoma who had related procedural claims.

Home Health Agency and Hospice data sets were queried for patients who died of melanoma. Melanoma-specific deaths were determined by the “vital status by COD” (where COD represents “cause of death”) variable in the Patient Entitlement and Diagnosis Summary File of the SEER-Medicare database. Variable values are as follows: 1, alive; 2, dead, cause not cancer; 3, dead, cause in situ, benign, or unknown behavior cancer; 4, dead, cause cancer; and 5, dead, cause unknown. We considered that patients who were diagnosed as having melanoma and who had a corresponding “4, dead, cause cancer” code in the Patient Entitlement and Diagnosis Summary File had died of their melanoma.

Because palliative care providers may bill by an ICD-9 code for symptom, complication, or metastases, we included these equivalents (Table 2). Our approach remained specific to claims related to melanoma and its sequelae.

We allocated charges to 3 time phases: initial, interim, and terminal. Classification of treatment costs by phases has precedence in the literature.^6,13 We determined the length of each phase that was most meaningful to our sample of patients with melanoma by observing trends in our data set as follows.

The end of the initial phase was defined as the point at which average monthly charges dropped by more than half, which was 4 months in our sample. We defined the beginning of the terminal phase as 6 months before death from melanoma. A 6-month terminal phase was selected to maximize the number of interim claims that could be obtained for patients diagnosed in advanced stages. The interim phase was defined as the remaining months separating the initial and terminal phases. This phase represents a relatively steady amount of care between the initial diagnostic evaluation and terminal care costs. The interim phase was applicable to patients surviving longer than the initial and terminal phases (10 months total) because the additional months of life were allocated to the interim phase. Because nearly half of the patients diagnosed as having stage IV melanoma died within 1 year, our selection of a 6-month terminal phase allowed for meaningful calculation of interim phase costs for patients with stage IV disease. For patients surviving less than 10 months, claims from the final 6 months were considered terminal and claims from any remaining months were considered initial.

We undertook a 2-step process to calculate annual direct cost. First, we used a cross-sectional approach, in which we combined phase charges from our data set with published information on the annual melanoma incidence and mortality in the US population 65 years or older. The second step was to convert charges to costs by using cost to charge ratios.

The first step is shown in Table 3. We calculated the incidence of melanoma and deaths by stage by incorporating published data. Specifically, we obtained the number of patients diagnosed in 1996 in the US population and the proportion of those who were 65 years or older.^14,15 The proportion of patients diagnosed in each stage of melanoma was determined from our sample data and was applied to estimate the number of patients 65 years or older diagnosed in each stage in the United States in 1996. From the same published data, we found that 4365 patients 65 years or older died of melanoma in 1996.^14,15

Categorizing deaths by stage is more complicated. Calculating the proportion of deaths by stage in a patient sample is potentially misleading because it depends on the number of patients in the sample initially diagnosed in each stage. For example, if we categorize the deaths observed in our sample (during the 6-year observation period by stage), we find that 5% of the deaths are attributable to stage 0, 25% to stage I, 28% to stage II, 28% to stage 3, and 14% to stage IV. Although 38 (70%) of our stage IV patients died within the observation period (Table 4), stage IV patients account for only 2.9% of our total patient sample (Table 4).

Our best approach was to categorize the deaths as attributable to stage IV disease, which is a method used in previously published work.¹⁶ However, this was done with one exception: from our data, we knew the proportion of patients in each stage who died within 1 year of their diagnosis. Accordingly, we had to appropriately assign the terminal costs from these patients to their respective diagnostic stages in considering their annual costs because there was no ambiguity in attributing these costs to specific stages.

Because we assigned deaths to stage IV disease as part of our method, and then calculated per-patient costs by stage, we assessed the variance of the per-patient stage IV costs if no patients (ie, no deaths) were assigned to stage IV.

Our phase charges were applied as follows. For patients who died of melanoma within 1 year, we included 4 months of initial, 2 months of interim, and 6 months of terminal phase charges. Although many of these patients may not have accumulated costs for the entire 12-month period, it was impossible to determine how many patients died at each time interval during the year. Thus, our calculation was a steady-state estimate. For patients who survived for more than a year, we included 4 months of initial and 8 months of interim phase charges. For patients who died of melanomas diagnosed in previous years as well as relapses, we assumed 6 months of terminal and interim charges.

To convert charges to costs, we used a cost to charge estimate. Formal cost to charge analyses require applying such ratios that are specific to hospitals, providers, and year. Rather than ascertaining and applying so many cost to charge ratios, we took a single estimated ratio to apply to our overall charges. According to the Carrier Reasonable Charge and Denial Activity Report from the Health Care Financing Administration (now Centers for Medicare and Medicaid Services), the national average of the ratio of reimbursements to submitted charges was 73.8% in 1985 (range, 66.2%-81.8%).¹⁷ Such annual cost to charge ratios were not available to us for the years of our data set (1991-1996), nor do we have access to hospital data to calculate an approximation of such a ratio. However, we have access to 5016 hospitals' cost to charge data for 2000: the average of these ratios was 0.539. Given that the median time between 1985 and 2000 falls in the time frame of our data set (1992), we estimated our single cost to charge ratio as the median between 0.738 and 0.539, which is 0.639.

We noticed that there were 263 patients who were diagnosed and died within the 6-year period that our data spanned. We considered that this select group of patients arguably had the greatest melanoma burden, given that they had a relatively short survival period and death was the uniform outcome. We calculated direct lifetime costs for this group to help further characterize the overall economic burden of melanoma, although the period of observation differed from the 1-year time frame required for estimating annual costs.

Demographic characteristics of the sample of 1858 patients are presented in Table 4.

Average monthly per-patient resource use by stage and phase is detailed in Table 5 and represented over time in the Figure. Consumption was greatest in the terminal phase ($3933) and did not vary significantly by stage, as expected, because all of these patients died of their disease, so any differences would be random. Consumption was less in the initial phase ($2194) and was least in the interim phase ($902). Initial- and interim-phase consumption was 3 to 5 times greater in stages II, III, and IV than in stage 0.

Table 3 gives the annual melanoma incidence and deaths by stage, calculated for the US population 65 years or older in 1996. The incidence and mortality by stage were combined with phase costs, shown in Table 6. The annual direct cost of treating melanoma in the Medicare population was estimated to be $249 million (Table 7). We calculated that stage IV annual costs were $25 019 per person if no patients (ie, no deaths) were assigned to that stage. This amount is comparable to the $23 285 per-person annual costs calculated by classifying deaths as stage IV disease (Table 7).

Fourteen percent of our sample (263 patients) died of melanoma during the 6-year observation period. This group, with the shortest survival, lived an average of 26 months after the diagnosis and incurred average per-patient melanoma costs of $28 210 ($44 148 in charges) from the time of diagnosis to death, or $13 020 ($20 376 in charges) per patient, per year.

Our study represents a population-based estimate of economic burden and direct costs of melanoma in Medicare beneficiaries 65 years or older. Previous studies have used population-based data to estimate costs for other malignant neoplasms, and we find the economic burden from melanoma to be comparable to that of other common malignant neoplasms in the elderly population. Early-stage melanoma costs appear similar to those of prostate cancer, whereas late-stage melanoma costs resemble those of colon cancer. Direct monthly per-patient costs of prostate cancer (all stages included) and stage II melanoma are comparable, at $1699¹⁸ and $1320 in 1996 dollars, respectively. For colon cancer, Brown et al⁶ found that patients with stage IV disease incurred lifetime, per-patient cancer costs of $29 400 (during a median 7-month survival). Our observed lifetime, per-patient costs of melanoma were $28 210 (during an average 26-month survival) for the 263 patients who were diagnosed as having melanoma and died of the disease between fiscal years 1991 and 1996.

When we compared monthly Medicare payments by phase for patients in all stages of melanoma with those for patients in all stages of colorectal cancer⁶ (Table 8), we found that melanoma was less costly than colorectal cancer in the initial phase ($1402 vs $3725 per patient). This is likely due to hospitalizations and complex surgeries in the initial management of colorectal cancer¹⁹ that do not usually occur for melanoma. However, melanoma costs were comparable to those of colorectal cancer in the terminal phase ($2513 vs $2259 per patient, respectively). The lack of definitive, effective therapy for melanoma, which may result in utilization of multiple chemotherapeutic agents in these later stages, likely drives up the costs.

A study by Tsao et al¹⁶ used microcosting to estimate direct melanoma costs for the general US population, including all age groups. The selection of our data set from fiscal years 1991 through 1996 was representative of a similar time frame,¹⁶ during which diagnostic and treatment modalities were consistent but specific only for patients 65 years or older. Because of the differences in demographics and methods, we found that a meaningful comparison of results was limited. For example, when we compared costs for stage III and IV melanoma, the Tsao et al model figured more than our estimates, which may have been because Tsao and coworkers' model was not limited to patients 65 years or older. The annual cost for stage III and IV according to the Tsao et al model was $40 735 and $41 459, respectively,¹⁶ vs $17 774 and $23 638 (Table 6) in our study (both in 1996 dollars). A possible explanation is the low utilization of interferon in our study sample, administered 44 times in total, allowing for a maximum of 23% of stage III patients to receive a single treatment. This differs from the Tsao et al model, which assumed that 80% of stage III patients received a full year of interferon therapy at a cost of $30 000.¹⁶ Perhaps not all interferon use was captured in the database, or there is underutilization in patients 65 years or older. This latter point would be in agreement with previous studies on chemotherapy²⁰ and radiation²¹ use in breast and rectal cancers, respectively, demonstrating that treatment declines significantly as the age of the patient increases.

Our study found that the annual direct costs were $249 million. Future work will be necessary to examine costs by category and to delineate expenditures for imaging and various treatment modalities, for example.

As a claims-based analysis, this study has several relevant limitations that may have underestimated resource consumption. First, we were unable to capture copayments and deductibles. Second, we lacked access to information on rationale for procedures, so claims attributed to melanoma may have been for unrelated conditions or the converse may have been true. Third, the data may have been incomplete. For example, as mentioned previously, not all interferon use may have been captured in the database. Fourth, elderly patients in the SEER regions tend to have lower rates of cancer mortality than elderly persons living outside of those areas.⁷ Another point of potential underestimation was in calculating direct annual costs, in which we were unable to account for melanomas diagnosed in previous years that did not result in death; patients whose data were not captured in incidence or death statistics during 1996 were not identified. Finally, to capture the true economic burden of melanoma from a societal perspective, we would need indirect and intangible costs.

A limitation of our study design may have resulted in an overestimation of melanoma-related costs. We made several assumptions for calculating a cost to charge ratio. We assumed a linear rate of decline between the cost to charge ratio from 1985 to 2000, although there most likely has been a faster decline. There is also the possibility that the costs of melanoma appeared more comparable to colorectal cancer costs before the approval of expensive drugs for treating colon cancer, such as bevacizumab and oxaliplatin. Of course, with the ongoing development of targeted melanoma treatments, our data may be a poor predictor of future costs.

Finally, there were some limitations related to our categorization by stage and phase. In terms of the phases, we used a single definition for initial, interim, and terminal time phases, although in reality these phases have varying relevance for different stages. For example, terminal-phase costs were calculated solely for patients who died of melanoma, and, while pertinent for stage IV disease, the terminal costs were not particularly relevant for in situ or stage I melanomas. However, the extremely small number of patients who were diagnosed as having in situ or stage I melanomas and then died of the disease was taken into account when the overall costs of melanoma were calculated. Patients diagnosed in these early stages simply contributed very little to the terminal-phase expenditures because they accounted for a minuscule proportion of these overall costs (Table 3). Furthermore, there are varying lengths of time of the interim phase, which may account for some cost variations by stage. We present the Figure to display generally that resource consumption (charges) by phase is consistent with expected trends, but this figure should be interpreted with caution: the resource consumption displayed is not a continuum; rather, it is displayed as discrete entities classified by phase. An example of the complexities involved in interpreting the Figure is that stage 0 charges appear to approach $4500 in the terminal phase, but, as discussed, it is important to remember that terminal phase costs were calculated for only about 4% of patients with stage 0 melanoma (Table 3) and that, when we calculated the total costs by stage, these terminal costs contributed very little.

We did have a major limitation in categorizing patients by stage when we expanded our data to the whole US population 65 years or older and had to classify patients in the entire United States who died of melanoma into stages. As discussed in the “Methods” section, we categorized these patients in stage IV. In our “Results” section, we discuss that the per-patient costs do not differ significantly after assigning deaths to stage IV disease ($25 019 vs $23 285 per year after assigning deaths to stage IV). However, when we interpret the total annual costs by stage (Table 7), it is likely that our total annual cost of stage IV disease ($99 004 678) might be a bit lower in reality, because some of the patients categorized as having stage IV disease since they died of melanoma may in fact have had stage III disease. In that case, we might expect our stage III total annual costs (representing all patients) to be a bit higher in reality than our estimated $28 249 054 total annual cost (Table 7).

Despite the common complexities encountered in economic analyses and the limitations specific to our methods and patient sample, our data demonstrated expected malignancy trends, lending face validity. Such trends include a U-shaped pattern of melanoma-related resource consumption over time¹⁹ and increasing consumption by stage, which are immune to any cost to charge ratio applied.

In conclusion, the large economic burden of melanoma in patients 65 years or older is comparable to that of other common malignant neoplasms in this population. Although relatively few patients actually die of this condition, the per-patient expenditures have particular significance in late-stage disease. The majority of consumption is attributable to advanced-stage disease and the terminal phase of treatment. If all patients were diagnosed and effectively treated in stage 0 or I, we estimate that the annual direct costs for the population 65 years or older would be between $99 million and $161 million, or 40% to 65% of their current value of $249 million. Policy guidelines for melanoma screening should consider that patients 65 years or older represent an increased risk, and, thus, economic burden, for later-stage melanoma.

Correspondence: Suephy C. Chen, MD, MS, Department of Dermatology, Emory University, 5001 Woodruff Memorial Bldg, 1639 Pierce Dr, Ste A-1100, Atlanta, GA 30322 (Schen2@emory.edu).

Accepted for Publication: November 12, 2009.

Author Contributions: Drs Seidler and Chen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Seidler, Pennie, Veledar, Culler, and Chen. Acquisition of data: Seidler and Pennie. Analysis and interpretation of data: Seidler, Pennie, Veledar, Culler, and Chen. Drafting of the manuscript: Seidler and Culler. Critical revision of the manuscript for important intellectual content: Pennie, Veledar, Culler, and Chen. Statistical analysis: Veledar and Culler. Obtained funding: Chen. Administrative, technical, and material support: Seidler. Study supervision: Chen.

Financial Disclosure: None reported.

Funding/Support: Dr Chen was supported in part by Mentored Patient Oriented Career Development Award K23AR02185-01A1 from the National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health, and an American Skin Association David Carter Martin Career Development Award.