Vestibular schwannomas (VSs; also known as acoustic neuromas) are benign tumors that represent approximately 10% of intracranial primary brain tumors.¹² Although they may be asymptomatic, they often present with unilateral hearing loss, tinnitus, imbalance, or vertigo. When they are discovered, VSs are managed in 3 ways: observation (the “wait and scan” approach); Gamma Knife surgery (GKS); or microsurgery. There is a significant body of literature (including many single- and multi-institution studies, as well as meta-analyses) regarding which management approach is superior.^10,11 One particularly nice study describes a decision-analytical model for determining which treatment option—observation, radiosurgery, or surgery—will produce the best quality of life outcome for a patient with VS.¹⁵

However, there is very little written about the cost of treating these patients. In a comprehensive PubMed search, we found only 4 studies that attempt to address the “cost” of VS surgery; however, these studies either did not use actual cost data or did not perform complete economic analyses.^6,13,14 In one study, Sonig et al. reported the average charges per hospitalization ($76,365.09 ± $58,039.93) for 2589 patients who underwent surgery for VSs between 2005 and 2009, as determined from the Nationwide Inpatient Sample (NIS) database.¹³ In the second paper, Verma et al. performed a cost comparison within the Canadian single-payer health care system and reported the average cost of treating a VS with microsurgery (CAD $22,402) versus LINAC radiosurgery (CAD $27,659) versus observation (CAD $9651).¹⁴ In the third paper, Banerjee et al. calculated initial and follow-up costs for 32 radiosurgery patients ($16,143) and 21 microsurgery patients ($23,788) at their institution (Mayo Clinic) between 2000 and 2002.¹ Finally, a British group published the first cost-utility analysis for VS treatment using estimated, as opposed to actual, costs of care in the United Kingdom.⁶

To date, no one has performed a cost-utility analysis with actual costs of care for patients with VS under conservative and nonconservative management paradigms in the multipayer system here in the US. We therefore performed a cost-utility analysis for the treatment of VSs at University of California at San Francisco (UCSF) between 2010 and 2013 under 3 different paradigms: observation, radiosurgery, and surgery.

Methods

We searched the UCSF neurosurgery database for all patients with the diagnosis of VS (ICD-9 code 225.1) who were treated at UCSF between 2010 and 2013. These were consecutively treated patients, and no patients were excluded. This time frame was selected because it is the period for which we have accessible UCSF cost data. After approval from our internal review board, we carefully reviewed the medical records to confirm the VS diagnosis, and then we divided up our patients into those who were observed, those who were treated with GKS, and those who received surgical treatment. Surgical patients had the diagnosis confirmed with pathological investigation, whereas patients in the GKS and observation groups had MRI findings consistent with VS.

Cost data were obtained from the UCSF hospital cost accounting database for our selected patient cohort. Both hospital charges and total costs (direct and indirect, as calculated in the UCSF financial accounting databases) were obtained for each separate patient encounter (i.e., hospitalization, office visit, and so on) and summed for each patient. The intent was to capture the entire cost of care associated with the VS, not just the index hospitalization. Averages are expressed as ± SD. Cost data manipulation was performed using Matlab scripts. Each inpatient cost item was classified into different categories: operating room ([OR] including OR time, anesthesia, and all equipment/supplies); hospital room; diagnostic tests/labs; imaging; medications/blood; physical therapy/occupational therapy/respiratory therapy (PT/OT/RT); and other. Each outpatient cost item was classified into separate categories: diagnostic tests/labs; imaging; medications/blood; radiation-oncology (including direct and indirect costs of running the Gamma Knife machine, as well as nonphysician personnel costs for radiation planning); and outpatient clinic. Of note, the hospital cost accounting database did not include physician fees, so the physician fees (specifically, physician reimbursements) for the neurosurgery, otolaryngology head and neck surgery, and radiation-oncology departments were obtained from separate databases and added to the total hospital costs. Of note, non-UCSF costs incurred in the care of a patient with VS (e.g., inpatient rehabilitation stay after discharge, outpatient PT/OT) are not available in our database and are therefore not included in our calculations.

We then created a decision-tree model in Microsoft Excel of the 2 most debilitating VS symptoms and worst outcomes (vertigo and death) to analyze the cost-effectiveness of our 3 treatment options by using outcomes information from the largest VS meta-analysis.¹⁵ Health utility indexes, which are preference-based measures of health-related quality of life (based on patients' ratings of the following domains: vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain), were taken from the published literature (0 for death, 0.62 for vertigo, 1 for perfect health).⁹ We selected these 2 outcomes because they are more disabling (i.e., associated with lower health utility index scores) than other symptoms of VSs, including facial paralysis (0.79)⁷ and unilateral hearing loss (0.82).³ We created separate models for each decade of diagnosis (20, 30, 40, 50, 60, or 70 years old). Average life expectancies, rounded to the nearest year, contingent on current age, were determined from Centers for Disease Control tables (approximately 59 years at age 20, 50 years at age 30, 40 years at age 40, 31 years at age 50, 23 years at age 60, 16 years at age 70).⁸

For patients in the observation group, we estimated a cost of $750 per year, based on our actual costs for one MRI study (approximately $500) and one clinic visit per year. For surgery patients, we estimated a total cost of $82,250, with approximately $80,000 for surgery (from our actual cost data) and $750 per year for 3 years of follow-up. Therefore, the estimated costs are not restricted to the index hospitalization during which the surgery occurred, but rather the entire care. This includes pre- and postoperative visits, imaging, and immediate readmissions for complications such as CSF leak and wound infection. For patients in the radiation group, we estimated a cost of $10,000 for radiation treatment (from our actual cost data) and $750 per year for 5 years of follow-up. Note that patients who underwent radiation need longer follow-up because patients are not immediately cured with radiosurgery, as they often are with surgery.

Quality-adjusted life years (QALYs) are equal to Time (years) * Health (i.e., Utility). However, to capture the “real time” preference that people have (valuing events in the present more than events in the future), we used a discounting factor of 0.03 and calculated the “net present value” (NPV) for QALYS by using the following Excel formula: NPV(0.03, Range of Utility Values)*(1 + 0.03)^0.5. Incremental cost-effectiveness ratios ([ICERs] = Δ Cost/Δ QALYs) were computed when appropriate.

Finally, sensitivity analyses were performed to evaluate the effect of several parameters on our model for patients diagnosed at age 40. One parameter was varied at a time. Parameters included chance of death after observation, surgery, and radiation, as well as the chance of developing vertigo after observation; ranges were chosen as the limits of the 95% confidence interval from the largest meta-analysis.¹⁵ The choices of surgery/radiation/observation after initial surgery, radiation, and observation were also varied, given all possible permutations. Finally, the costs of observation, radiation, and surgery were varied ± 50%.

Results

Decision-Tree Model of Patients With VS

Figure 1 demonstrates our decision-tree model for the 3 basic treatment options for VSs (observation, radiosurgery, and surgery). The chance frequencies for vertigo and death after each treatment option are derived from the largest VS decision-modeling meta-analysis.¹⁵ More specifically, observation comes with the highest chance of death (2.4%, vs 1% with radiation and 0.4% with surgery). Similarly, observation is associated with the highest rates of vertigo in the patients with VS (22%), versus 5.7% in patients in the radiosurgery group and 0.8% in patients in the surgery group.¹⁵

Decision-tree model of VS treatment options. Squares indicate decision nodes, circles are chance nodes. Chance frequencies are derived from Whitmore et al., J Neurosurg 114:400–413, 2011.

Discussion

Our study is one of the only reports on the total cost of treating VS patients in the US. Our surgical costs (average $80,074) are higher than the only other reported costs from the US ($23,788 for surgery at Mayo Clinic between 2000 and 2002¹), which may reflect institutional differences as well as the rising cost of health care over the past decade. Interestingly, however, our average radiosurgery costs ($9737) are lower than those reported in the Mayo Clinic study ($16,413), perhaps due to decreasing radiosurgery costs with improving technology over time. Similarly, our surgical costs in the multipayer system here in the US are higher than those reported in Canada ($80,074 in our study vs $22,402 in Canada¹⁴), which is consistent with other studies showing similar cost discrepancies between the 2 countries for cardiac bypass surgery.⁴

One of the strengths of our study is that we calculate hospital costs, as compared with charges (i.e., what the hospital bills the insurance company), which were reported in the study by Sonig et al. in which the NIS database was used.¹³ Several recent studies show that hospital charges (or billing) bear little resemblance to economic cost,² and use of hospital charges as a proxy for cost may lead researchers to draw unwarranted conclusions.⁵ The best measure of cost is actual resource utilization,⁵ which can be extremely difficult to calculate. We believe that by combining our direct cost data from the hospital accounting database with actual physician reimbursements, we have calculated the closest and most accurate proxy for the true cost of care for patients with VS that has been published to date. Although we appreciate that physician reimbursements may not accurately reflect the true “cost” of physician services, we believe that this is a better proxy than physician charges, which are often significantly inflated (to account for low reimbursement rates, for instance). Moreover, the physician reimbursement reflects the actual cost to society (in terms of dollars transferred in our economy) to pay for that physician's services. We acknowledge that a homogeneous method of calculation of cost for hospital-related care and physician services is needed so that we can better compare results across studies and organizations.

Furthermore, we provide the first published breakdown of hospital costs for these patients. In our current economic climate, price transparency is one of the first, and necessary, steps toward cost containment. As we would expect, we find that the OR and hospital stay are the most expensive aspects of hospitalization for a patient with VS. More specifically, 1 minute of OR time costs between $53 and $78, an ICU bed costs $4337 per day, and a floor bed costs $1519 per day at our institution. This suggests that reduction of OR costs and length of hospital stay could significantly impact surgical costs. For example, just a 10% reduction in surgical costs (from $82,250 to $74,025) results in improved ICERs and makes surgery a cost-effective option for patients who are diagnosed at age 50 (ICER $143,255).

Although we recognize the limitations of our relatively small data set, we believe that our data and conclusions are more powerful than larger studies, such as the one by Sonig et al.,¹³ because of the accuracy of our cost data. Expanding our study to other institutions might help increase generalizability, but would also decrease confidence in our cost data and therefore introduce further sources of error. Unfortunately, none of the nationally available databases (such as the NIS or Premier databases) provide real cost data for this type of analysis. Another important limitation is our retrospective study design, and the different characteristics of our surgical, radiation, and observation groups. Specifically, our surgical group is composed of younger patients with larger, more symptomatic tumors. We acknowledge that a prospective study of all patients in whom VSs were diagnosed, controlling for tumor size, patient age, and symptoms, would provide the best measure of cost over a variety of treatment paradigms. Such a study would also enable us to determine treatment recommendations based on age, presentation, and tumor size (for example, what is the most cost-effective treatment for a 40-year-old patient who presents with a < 1-cm asymptomatic VS?), detailed questions that we cannot fully answer with our current analysis.

Finally, our cost-utility model provides an important guideline for neurosurgical practice. Unlike the only other VS cost-utility analysis in this area,⁶ our model calculates utilities based on the chances of vertigo and death. We have selected these 2 outcomes (as opposed to the hearing loss, facial nerve dysfunction, and so on examined in Gait et al.'s study) because these are 2 of the most debilitating symptoms/complications of VSs. Patients can often continue to have a relatively normal life with unilateral hearing loss, but symptoms of vertigo that lead to dizziness, nausea, and imbalance can have a significant impact on their well-being and ability to work. These differences are reflected in published health utilities, which, for example, are lower for vertigo than for unilateral hearing loss or facial paralysis.⁹ We acknowledge that selection of these 2 outcome metrics may bias our results in favor of surgery, given that facial paralysis and hearing loss outcomes are more favorable for radiosurgery than for open surgery.

Another major improvement in our model is that we consider many decades of diagnosis (from 20 to 70 years old), whereas Gait et al. only looked at a starting age of 58 years.⁶ Our longer duration of follow-up probably explains the difference in our results, in that we find that surgery is a cost-effective option for patients < 45 years old. Lastly, we understand that patients may have staged treatments (planned surgery, followed by radiation) or may have multiple rounds of either surgery or radiation, which are not accounted for in our model. Similarly, there are many additional options, including return to emergency room, readmission for complication, and so on that we have not included. We have assumed only 2 decision points in our model to avoid this extraordinary complexity.

Conclusions

We report that the average total hospital cost over a 3-year period for surgically treated patients with VS was $80,074 (± $49,678) versus $9737 (± $5522) for patients receiving radiosurgery and $1746 (± $2792) for patients who were observed. When modeling the most debilitating symptoms/complications of VSs (vertigo and death) at different ages at diagnosis, we find that surgery is a costeffective alternative to radiation when VS is diagnosed in patients who are < 45 years old. For patients ≥ 45 years, radiation is the most cost-effective treatment option for VSs. This is the first cost-utility analysis for VSs using real cost data published in the US.

Disclosure

Matthew Sun was a Howard Hughes Medical Institute Medical Research Fellow. Dr. Parsa is partially funded by the Michael J. Marchese Chair in Neurosurgery. The authors declare no conflict of interest.

Author contributions to the study and manuscript preparation include the following. Conception and design: Zygourakis, Barani, Parsa. Acquisition of data: Zygourakis, Oh, Sun. Analysis and interpretation of data: Zygourakis, Barani, Kahn. Drafting the article: Zygourakis. Critically revising the article: Zygourakis, Oh, Sun, Barani, Kahn. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Zygourakis. Statistical analysis: Zygourakis, Kahn. Study supervision: Zygourakis, Parsa.