Malignant Melanoma: The Implications of Cost for Stakeholder Innovation
An activity-based cost accounting model of melanoma care was used to identify key cost drivers and opportunities for stakeholder innovation.
The incidence of malignant melanoma (MM) increased from 7.9 per 100,000 in 1975 to 21.5 in 2007.1 In 2008, the Surveillance, Epidemiology and End Results (SEER) data showed that an estimated 62,480 cases of MM and 54,020 cases of melanoma in situ (MIS) were diagnosed in the United States.2
Although most cases are caught early, melanoma is devastating when diagnosed at a later stage. Five-year survival rates are 98.7% for localized disease but only 15.5% for distant disease.2 Moreover, the median age at diagnosis is between 45 to 55 years, coinciding with peak professional productivity and family responsibility. On average, melanoma decreases life span by 17.1 years.3
Because of this effect on society, investigators have quantified melanoma’s economic impact. Tsao et al4 estimated the total 1997 direct cost of melanoma to be $563 million. Seidler et al5 calculated the 1996 economic burden in the elderly to be $390 million, while Alexandrescu6 compared the 2009 costs of treating each stage of disease. Together, these analyses emphasized calculating cost rather than practical implications of their findings.
We applied a managerial perspective to melanoma care and chose a cost accounting analysis to identify the specific are most important in controlling costs, prioritize key areas for innovation, and recommend ways forward for entrepreneurs, investors, policy makers, and managers.
MATERIALS AND METHODS
We created a survival-adjusted activity-based cost accounting model for MM and MIS, using the clinical practice guidelines published by the National Comprehensive Cancer Network (NCCN) v2.2008.7 Adjusting these guidelines based on reported trends in the literature and our clinical experience, we performed sensitivity analyses to determine whether our assumptions substantially affected the model.
illustrates the clinical care activities included in our analysis. Key assumptions are shown as a percentage of treated patients. Estimates of patient survival by stage were taken from the American Joint Committee on Cancer (AJCC) Melanoma Database.8 Recurrent melanoma survival data came from the literature.9
Our ground-up approach accounted for all costs borne by payers (public or private) based on 2008 national average Medicare Current Procedural Terminology (CPT) codes, negotiated drug therapy, and diagnosis-related group charges. For physician care, CPT codes were reimbursed at 80% of Medicare, consistent with practice management data presented at the 2007 American Academy of Dermatology meeting.10 Reimbursement for procedure-based CPT codes were further adjusted by the formula: 100%, 50%, and 25% for the fi rst, second, and third CPT codes, respectively. Diagnosis-related group charges were calculated for Massachusetts General Hospital using the Centers for Medicare & Medicaid Services PPS Inpatient PC Pricer program and allocated for each admission using the average length of stay.11
Drug therapy costs were discounted 15% off 2008 average wholesale price,12-14 reflecting a 2002 report by the Office of the Inspector General that showed pharmacy acquisition prices ranging from 17.2% to 72.1% below average wholesale price15 and Medicaid’s routine application of 5% to 17% discounts off average wholesale price for branded drugs.13 We did not explicitly address drug assistance programs.
We estimated 2008 clinical trial enrollment to be 3%, 20%, and 64% of newly diagnosed stage II, III, and IV patients from the National Cancer Institute clinical trial database (phase II and III, treatment trials only), respectively.16 Our research experience indicates that imaging and laboratory testing are typically covered by payers. Because recurrent cases are usually eligible for trials that recruit stage IV patients, we allocated enrollment evenly across both groups, reducing the participation rate to 15%. Our analysis does not include indirect costs (eg, lost productivity).
We assumed that melanoma care began with a new patient exam by a dermatologist (CPT 99204)17 and included a biopsy if the lesion was suspicious for cancer (CPT 11401) and pathology examination (CPT 88305). We included cases that yielded only a positive diagnosis and adjusted 4% for inadequate biopsies that needed reexcision.18
Stage 0 melanoma is excised at a follow-up appointment and evaluated by a pathologist. Charges vary based on size of the excision; the NCCN recommended excision margin is 0.5 cm; follow-up (CPT 99214)17 is every 12 months.
Stage IA care includes follow-up every 3 months for 2 years, every 6 months for 1 year, and every 12 months thereafter; excision margins are 1.0 cm. NCCN guidelines distinguish “stage IA with adverse features,” patients with fatigue or a biopsy with adverse features (eg, mitosis, dedifferentiation). We incorporated these patients’ care as stage IB/IIA.
Patients with stage IB to IIC require greater surveillance. Because lesions are larger, excision may require a flap or graft. To stage care, 75% of stage IB/IIA and 90% of stage IIB/IIC patients undergo sentinel lymph node biopsy (SLNB).19 Use of this procedure is supported by reports indicating that 11% of melanomas with vertical growth of less than 2 mm may be SLNB+.20,21 NCCN recommends imaging (eg, chest x-ray [CXR]) if clinically indicated.7 While studies have shown the low clinical utility of CXR,22,23 New York University’s experience between 2002 and 2005 with stage IB and II disease showed that 80% of stage IB and II patients received preoperative CXRs and at least 43% received computed tomography (CT) scans prior to SLNB.24 We excluded outpatient lactate dehydrogenase and complete blood count testing because they are not justifi ed,25 except in stage IV patients.
The decision to undergo interferon (IFN) alfa-2b therapy, costing $39,013 per patient, is important. We assumed all patients received high-dose IFN (20 μg, 5 doses per week for a month and 10 μg, 3 doses per week for 11 months) described in the Eastern Cooperative Oncology Group and Intergroup Trials and were treated for side effects (eg, with metoclopramide, selective serotonin reuptake inhibitor/mirtazapine, and megestrol acetate).26 We included monitoring (eg, CXR, electrocardiogram, blood) and oncologist follow-up schedules from these studies.
Stage 3 disease has many options for treatment.27 Patients with a positive SLNB are further evaluated by CT and offered lymphadenectomy. Next, patients undergo a clinical trial, follow-up with imaging, or IFN therapy. We incorporated an 18% false-negative rate of SLNB for patients with clinically positive, in transit, and distant metastases. 28 If nodes are clinically positive, fi ne-needle aspiration (FNA) is performed. If inconclusive, nodal biopsy is performed. Primary wide excision, fl ap, or graft with lymphadenectomy is often the next step. Patients are then offered either follow-up with CT imaging, clinical trial, or IFN. Stage IIIC patients may also elect radiation.
Stage III in transit typically requires biopsy of a metastasized lesion to confi rm the diagnosis, followed by excision. After surgery, patients are offered hyperthermic infusion, intralesional injection, laser ablation, topical imiquimod, and/or systemic chemotherapy. Chemotherapy is often chosen, given the potential for remission. With multiple options, we chose monotherapy with dacarbazine (250 mg/m2 daily over 5 days) with 6 cycles and included metoclopramide for nausea because no other regimen had demonstrated a survival benefit.29 We estimated that 75% of patients were treated at outpatient facilities. 30 Five percent of patients undergoing chemotherapy were expected to undergo combination therapy with interleukin-2.31 We also included monitoring (eg, echocardiogram, CXR, electrocardiogram, pulmonary function tests, blood) and oncologist follow-up. After treatment, patients entered clinical trials, pursued watchful waiting, or started IFN.
For new stage IV lesions, FNA or image-guided biopsy with lab work and imaging assesses organ involvement and candidacy for surgery. If disease is limited (ie, resectable), then surgery is indicated followed by IFN, clinical trial, or follow-up. If the patient refuses or cannot tolerate surgery, chemotherapy or follow-up is indicated. Stage IV patients undergo resection in only 35% of cases.32 If MM is disseminated (ie, unresectable), the brain is evaluated and if involved, clinical trial, chemotherapy, or palliative care (surgery, radiation therapy, or supportive care) is indicated. Otherwise, successive rounds of chemotherapy are recommended if tolerated.
We assumed all patients receive palliative care, costing $29,872 in 2008. This estimate was derived from the 6-month 2001 average cost of palliative cancer care.33 To update these costs, we adjusted by the consumer price index.34
Recurrence may be local, regional, or distant. Patients with local and in transit metastases receive surgical care followed by stage IV therapy. Nodal and distant metastases receive FNA or nodal biopsy followed by stage IV treatment. Based on published experience, median disease- free interval (the period from the date of primary diagnosis to diagnosed recurrence) was 2.7 years (range, 0.37-27.6 years) for melanoma.9 We treated recurrent cases aggressively and applied stage IV costs, since even stage I and II patients with local recurrence (defined as
within 3 cm from the primary lesion) have survival rates comparable to those of patients with stage III disease (analysis of source data from Francken et al9 with permission; Figure 1). Although survival varies by site of recurrence, we used mean reported 1-, 2-, 5-, and 10-year survival rates of 56%, 40%, 20%, and 8.3%, respectively.9
To estimate total recurrent cases, we assumed that deaths comprised new patients in stages I through III whose disease had recurred and progressed rapidly, new stage IV patients, and new recurrent cases that were diagnosed in previous years.
demonstrates how this framework, when applied to the 2008 estimate of 8420 deaths, resulted in 15,256 treated cases of recurrent melanoma. To derive the number of survivors with recurrent disease (Table 2A, lines D and E), we used published recurrence survival curves.9
To estimate the contribution from the previous 10 years, we used 2003-2007 SEER incidence data and applied the techniques described above for distributing newly diagnosed patients into stages. SEER altered melanoma reporting in 2003. Costs of diagnoses prior to 2003 were estimated from our model (0.1% to 5.3% of total expenses, depending on stage). While guidelines advise lifelong follow-up, this follow-up may be deferred as more time passes since a positive finding. We assumed all patients attended the fi rst annual checkup and adjusted attendance by 10% annually thereafter (Goldsberry G, Kimball AE. Wait times in clinic practice: an operational model of outpatient care. Unpublished manuscript. Massachusetts General Hospital, 2009).
Table 2B summarizes how we estimated the mix of new MM. We distributed the 2008 SEER incidence estimates2 across stages I through IV based on the AJCC experience.8 Because medical care is not always unique to a pathologic stage (eg, T2a vs T2b), we combined pathologic stages into “care process” groups that were determined by NCCN guidelines. For example, pathology stage IA corresponded to care process stage IA. The stage III SLNB+ care group was derived from pathology stages IIIA and IIIB with micrometastases only and included 50% of stage IIIC N3 lesions because of the potential for micrometastases. (This pathology stage includes lesions of any thickness and ulceration with 4 or more positive nodes with micro or macro metastasis.) The stage III clinically positive nodes group included pathology stage IIIB patients with macrometastases only and remaining stage IIIC patients. Care process stage III in transit refl ected pathology stage IV M1a patients. Stage IV patients comprised the remainder.
Because data are commonly reported by clinical stage, we further grouped care process stages into clinical stages I through IV. Thus, 2008 SEER incidence fi gures were allocated: 32,571 in stage I, 20,373 in stage II, 6060 in stage III, and 3475 in stage IV.
We adjusted for patients initially diagnosed with and treated for stages II and III disease whose disease recurred within the same year, approximately 3882 cases in 2008. These patients were assumed to receive 6 months of care under the primary diagnosis and the remaining 6 months under stage IV recurrent cost burden.
The estimated 2008 cost of MM was $1563 million (
). The majority of costs were incurred by recently diagnosed patients, ranging from 74% to 91%. Although stage IV patients incurred the majority of costs (57%), that was primarily due to cases of recurrent disease (44%). Patients with stages I and II disease were responsible for 10% and 17% of the costs, respectively.
Not surprisingly, costs were higher with later stages and more recent diagnosis (Table 4). For example, the average patient diagnosed in 2004 with stage II disease could expect to incur only $35 in costs in 2008, compared with $12,556 if diagnosed in 2008. Stage IV melanoma costs are 40 times greater than those for MIS in the fi rst year. Clinical decisions can drastically affect the cost of care, especially in the later stages, as shown in Table 5. For example, stage IV costs ranged from $3462 to $156,374 for an individual.
highlights the estimated fi nancial impact of adjustments to some of our key clinical assumptions. For example, if palliative care costs were increased by $1000, the overall 2008 cost of care would rise $6.7 million. Conversely, decreasing IFN use by 10% in stage III and IV patients would reduce annual costs by approximately $17.7 million and $40.2 million, respectively.
Tsao et al4 published the fi rst estimate of the annual direct cost of MM. Since then, annual costs have increased from $563 million in 1997 to $1563 million in 2008. This analysis highlights the key drivers of this change.
With little innovation, cost increases have been driven by incidence and the consumer price index. Incidence alone increased 55% between 1997 and 2008, more than 22,000 cases annually, while the consumer price index increased 34%.34 Together, the consumer price index and the incidence have contributed $610 million to the annual costs, raising 1997 estimates to $1.17 billion.
Imaging, traditionally linked to escalating healthcare costs, contributed only $75 million, or 4.8% of 2008 mean annual costs. In fact, use of CT and CXR in melanoma reflects a rift between the academic literature and practice that has been reported in other areas of medicine.35 For more than a decade, the literature has shown that imaging early-stage disease is rarely useful.22,23,36-39 Yet, a 2007 New York University report demonstrated that imaging is still routinely ordered.24
Drugs, while a significant proportion of total costs, have not been responsible for cost increases. For example, IFN alfa-2b comprises 17% of melanoma costs, yet contributed only $41 to $49 million, or 2.7% to 3.1% (depending on discounts) of the increase in costs after infl ation between 199840 and 2008. Interestingly, IFN treatment costs, including physician visits and follow-up, have remained stable and may refl ect the impact of CPT codes not rising proportionally over time. Despite this legacy, drug costs should take center stage as use of ipilimumab and vemurafenib, whose treatment costs exceed $100,000 annually, grows after recent US Food and Drug Administration approval.41,42 This upcoming shift allows our analysis to set a baseline against which future researchers can compare costs.
Our remaining differences in cost between 1997 and 2008, approximately $380 million, refl ect shifts in melanoma incidence data, higher estimates of recurrent melanoma, and a broader cost perspective. The number of stage II and IV diagnoses increased by 1.8-fold and 2.1- fold, respectively.4 Similarly, published AJCC data increased recurrent melanoma estimates by 6421 after adjustment for incidence.9 Our comprehensive approach to melanoma care included surgery for metastases, clinical trials, and ambulance transport. Yet we did not include patient bad debt and claim rejections, the latter varying from 7% to 14%.40
Other modeled costs actually decreased. We used average national reimbursement levels and reduced palliative care costs to $29,872, lower in real terms than the Tsao et al 1997 $27,000 estimate4 and higher than Alexandrescu’s arbitrary $14,500 in 2009.6,43 Although home health use44 and care intensity33 can impact costs, our estimate reflects the most recent published Medicare data.
Our analysis also compares favorably with cost relationships reported from Medicare claims data, which permit comparison of melanoma costs with the costs of other cancers (eg, lung, colorectal).45 For many cancers, local disease is more expensive, perhaps refl ecting use of operating rooms to excise lesions rather than clinic rooms. In contrast, end-of-life costs in melanoma are among the lowest, presumably refl ecting rapid progression of advanced disease.
Any economic analysis depends on the assumptions used to build the model. Not everyone treats melanoma in the same way, and few data detail actual practice patterns in the community. To incorporate divergent perspectives, we calculated the impact of adjusting key variables (
) and believe that discordant views on care intensity should fall within 10% to 15% of our estimate. Our standard for this model is decision-making ability rather than certainty.
Understanding these cost relationships, managers have several tools to shape costs through adoption and reinforcement of evidence-based practice. For example, clinician updates can highlight disappointing results from the Sunbelt Melanoma Trial46 to dampen use of IFN while educating physicians about use of new therapies such as ipilimumab and vemurafenib. Electronic medical systems allow tracking the utilization of expensive tests and therapies, and uncover new insights into the care dynamics between specialties (eg, inappropriate imaging at New York University).24 By uncovering how and by whom clinical decisions are made, managers can target their efforts to shift traditional physician practice.
Managers will have to address patient management expenses arising from rapid increases in early-stage disease, 47-50 with costs growing from 10% in 19974 to 27% in 2008, by promoting technologies that risk-stratify lesions and predict treatment response. Haqq et al51 have demonstrated that gene expression signatures can predict melanoma progression and therapy response. By refi ning this and similar approaches, managers may ultimately help physicians prioritize treatments and minimize costs from trial-and-error approaches to therapy.
Illustrating the impact clinical decisions have on costs,
demonstrates that human-intensive activities like palliative care, diagnosis, follow-up, and excision represent 67% of annual costs. The signifi cant role played by low-tech activities in melanoma represents an opportunity for innovators and managers; they may be able to either disrupt these activities with novel solutions or refi ne them with evidence-based practice innovations.
Palliative care, approximately 20% of annual melanoma costs, has become a signifi cant problem for stakeholders. One approach has been to shift patient care to the home and incorporate remote monitoring to enable rapid response to potential problems. For example, Reid et al52 demonstrated 29% fewer emergency department visits and 6% fewer hospitalizations after implementation of the medical home model over 2 years. While hurdles regarding reimbursement and infrastructure remain, entrepreneurs with solutions in this space could have tremendous societal impact and personal returns.
Patients with treated and undetectable disease that recurs represent the largest cost burden, totaling $694 million or 44% of total costs in 2008. While dismal patient prognoses already encourage clinicians to seek new ways to prevent recurrence, we hope this analysis defi nes the scope of the problem for entrepreneurs and investors to focus their creativity and capital. Delaying the mean time to recurrence by 1 year saves the healthcare system $45 million annually.
In a resource-constrained healthcare system that spent $98 billion on cancer in 2008, economic analyses help stakeholders focus their efforts on improving patient care.53 Unfortunately, innovation must contend with capitated and rigid payment systems that often reduce payouts with improved outcomes.54 Despite these market distortions, the costs of managing early-stage melanoma, palliative care, and recurrence will grow. We hope this analysis will drive cost-effective management of melanoma and inspire innovations that improve care.