Payer Costs for Inpatient Treatment of Pathologic Fracture, Surgery to Bone, and Spinal Cord Compression Among Patients with Multiple Myeloma or Bone Metastasis Secondary to Prostate or Breast Cancer

BACKGROUND
Patients with bone metastasis secondary to prostate or breast cancer or multiple myeloma are predisposed to skeletal-related events (SREs), such as surgery or radiation to the bone, pathologic fracture, and spinal cord compression. Inpatient costs of these and other SREs represent an estimated 49%-59% of total costs related to SREs. However, information on payer costs for hospitalizations associated with SREs is limited, especially for costs associated with specific SREs by tumor type.


OBJECTIVE
To examine costs from a payer perspective for SRE-associated hospitalizations among patients with multiple myeloma or bone metastasis secondary to prostate or breast cancer.


METHODS
Patients with SRE hospitalizations were selected from the MarketScan commercial and Medicare databases (January 1, 2003, through June 30, 2009). Sampled patients had at least 2 medical claims with primary or secondary ICD-9-CM diagnosis codes for prostate cancer, breast cancer, or multiple myeloma and at least 1 subsequent hospitalization with principal diagnosis or procedure codes indicating bone surgery, pathologic fracture, or spinal cord compression. For patients with prostate cancer or breast cancer, a diagnosis code for bone metastasis was also required. If secondary diagnoses or procedure codes for SREs were present in the claim, they were used to more precisely identify the type of SRE for which the patient was treated, resulting in 3 mutually exclusive categories: spinal cord compression with or without pathologic fracture and/or surgery to the bone; pathologic fracture with or without surgery to the bone; and only surgery to the bone. Related readmissions within 30 days of a previous SRE-associated hospitalization date of discharge were excluded to minimize the risk of underestimating costs. Mean health plan payments per hospitalization, measured as net reimbursed amounts paid by the health plan to a hospital after subtracting patient copayments and deductibles, were analyzed by cancer type and type of SRE.


RESULTS
A total of 555 patients contributed 572 hospitalizations that met the study criteria for prostate cancer, 1,413 patients contributed 1,542 hospitalizations for breast cancer, and 1,361 patients contributed 1,495 hospitalizations for multiple myeloma. The mean age range was 61 to 72 years, and the mean length of stay per admission was 5.9 to 11.6 days across the 3 tumor types. The ranges of mean health plan payment per hospital admission across tumor types were $43,691-$59,854 for spinal cord compression, with or without pathologic fracture and/or surgery to the bone; $22,390-$26,936 for pathologic fracture without spinal cord compression, with or without surgery to the bone; and $31,016-$42,094 for surgery to the bone without pathologic fracture or spinal cord compression.


CONCLUSIONS
The inpatient costs associated with treating SREs are significant from a payer perspective. Our study used a systematic process for patient selection and mutually exclusive categorization by SRE type and provides a per episode estimate of the inpatient financial impact of cancer related SREs assessed in this study from a third-party payer perspective.

METHODS: Patients with SRE hospitalizations were selected from the MarketScan commercial and Medicare databases (January 1, 2003, through June 30, 2009). Sampled patients had at least 2 medical claims with primary or secondary ICD-9-CM diagnosis codes for prostate cancer, breast cancer, or multiple myeloma and at least 1 subsequent hospitalization with principal diagnosis or procedure codes indicating bone surgery, pathologic fracture, or spinal cord compression. For patients with prostate cancer or breast cancer, a diagnosis code for bone metastasis was also required. If secondary diagnoses or procedure codes for SREs were present in the claim, they were used to more precisely identify the type of SRE for which the patient was treated, resulting in 3 mutually exclusive categories: spinal cord compression with or without pathologic fracture and/or surgery to the bone; pathologic fracture with or without surgery to the bone; and only surgery to the bone. Related readmissions within 30 days of a previous SRE-associated hospitalization date of discharge were excluded to minimize the risk of underestimating costs. Mean health plan payments per hospitalization, measured as net reimbursed amounts paid by the health plan to a hospital after subtracting patient copayments and deductibles, were analyzed by cancer type and type of SRE.
RESULTS: A total of 555 patients contributed 572 hospitalizations that met the study criteria for prostate cancer, 1,413 patients contributed 1,542 hospitalizations for breast cancer, and 1,361 patients contributed 1,495 hospitalizations for multiple myeloma. The mean age range was 61 to 72 years, and the mean length of stay per admission was 5.9 to 11.6 days across the 3 tumor types. The ranges of mean health plan payment per hospital admission across tumor types were $43,691-$59,854 for spinal cord compression, with or without pathologic fracture and/or surgery to the bone; $22,390-$26,936 for pathologic fracture without spinal cord compression, with or without surgery to the bone; and $31,016-$42,094 for surgery to the bone without pathologic fracture or spinal cord compression.
CONCLUSIONS: The inpatient costs associated with treating SREs are significant from a payer perspective. Our study used a systematic process for

R E S E A R C H
• The current study reports costs for inpatient treatment of SREs from the payer perspective, measured as net payment to the facility after subtracting patient copayments and deductibles, by type of SRE across 3 tumor types: prostate cancer, breast cancer, and multiple myeloma. • Mean inpatient cost estimates for the 3  B one is a common site for the spread of several malignancies as its nutrient-rich environment provides a favorable soil for colonizing tumor cells. 1,2 When tumor cells metastasize to bone, they are thought to secrete cytokines and growth factors that induce osteoblasts (bone-forming cells) to release the protein RANK ligand (RANKL). In turn, RANKL promotes the formation and survival of osteoclasts (bone-resorbing cells), which, when activated, cause local bone destruction in the direct area of the tumor metastasis. Key growth factors are released from the bone breakdown that may promote proliferation, metastasis, and survival of tumor cells. Thus, a "vicious cycle" of tumor expansion and bone destruction resorption is perpetuated. 3 Bone metastases can result in significant skeletal complications known as skeletalrelated events (SREs), such as pathologic fracture, spinal cord compression, or need for surgery or radiation to bone. [4][5][6][7][8] Nearly 70% of patients with metastatic prostate or breast cancer experience metastases to the bone, and up to 95% of patients with multiple myeloma experience osteolytic bone lesions, which may lead to SREs. [9][10][11][12] Breast and prostate cancers are the most prevalent cancers in the United States, with 1 in 8 women developing breast cancer and 1 in 6 men developing prostate cancer during their lifetimes. 13 Additionally, virtually all cases of myeloma involve bone destruction. 14 Given these facts, the economic burden that SREs associated with these cancers can place on health systems may be substantial. Treatments that can reduce or prevent SREs may reduce the economic burden of these skeletal complications. Intravenous bisphosphonates, primarily zoledronic acid (Zometa; Novartis), are effective for preventing SREs. 15 Denosumab (Amgen), a RANKL inhibitor, has effectively delayed and reduced the occurrence of SREs compared with zoledronic acid in ongoing clinical trials. [16][17][18] To better assess the potential cost-offsets gained by these therapies, it can be helpful to understand the cost burden that each type of SRE places on the health care system.
Treatment of SREs occurs in both inpatient and outpatient settings. Inpatient treatment represented an estimated 49%-59% of total SRE costs in previous research. 19,20 This analysis focuses on third-party payer costs for treating SRE episodes in the inpatient setting by type of SRE. Previously published cost-• As the focus of this study was on inpatient payer costs of SREassociated treatment, radiation to bone was not included as an independent SRE. • The results of this study will be the first published inpatient cost estimates for SREs in patients with multiple myeloma in the United States.

Sample Selection
The study sample consisted of patients with multiple myeloma, prostate cancer, or breast cancer with claims between January In addition, at least 1 non-rule-out bone metastasis diagnosis (ICD-9-CM 198.5) before or on the date of admission was required for patients with prostate cancer or breast cancer. SRE hospitalizations were further categorized using  secondary diagnosis (ICD-9-CM) and/or secondary procedure (ICD-9-CM or CPT) codes indicative of local irreversible events defining an SRE as selected by a third-party physician with expertise in ICD-9-CM and CPT coding ( Figure 1). Diagnosis codes identified pathologic fracture, surgery to bone, or spinal cord compression, and procedure codes identified surgery to bone or spinal cord compression (Appendix). Although radiation to bone is also a local irreversible event, it was not included as an independent SRE in this study of inpatient costs because the procedure is primarily conducted in the outpatient setting. Hypercalcemia of malignancy, another recognized complication of bone metastases, is a systemic and potentially reversible event and was thus not considered to be a component of the SRE. 21 Bone pain was similarly not included. Hospitalizations with negative or no reimbursed amounts were excluded from the analysis. In addition, hospitalizations within 30 days of a previous SRE hospitalization discharge date were excluded to avoid counting readmissions resulting from complications from the previous hospitalization.

SRE Identification and Categorization
When secondary diagnosis or procedure codes for SREs were present in the claim, they were used to more precisely identify the type of SRE for which the patient was treated, as we assumed that, on average, 1 SRE per admission occurred and that the procedure was performed to treat the diagnosis   identified in the claim. For instance, surgery to bone includes procedures to set or stabilize a fracture or to prevent an imminent fracture or spinal cord compression. We verified our assumption in the breast cancer sample and found that, for all claims with spinal cord compression and surgery codes, the surgery location was to the spine. Similarly, for all claims with spinal cord compression and pathologic fracture codes, the location of the fracture was the vertebrae. For claims with pathologic fracture and surgery codes, the surgery site was consistent with the fracture site in all but 3% of cases.
Consequently, in cases where both surgery and pathologic fracture codes were present, we assumed that the surgery was to treat the pathologic fracture, and thus the SRE was classified as a pathologic fracture. In cases where both surgery and spinal cord compression codes were present, we assumed that surgery was to the spine and thus categorized the SRE as a spinal cord compression. In cases where both pathologic fracture and spinal cord compression codes were present, we assumed that the location of the fracture was the vertebrae and thus categorized the SRE as a spinal cord compression. This method created 3 mutually exclusive categories: spinal cord compression with or without pathologic fracture and/or surgery to the bone; pathologic fracture with or without surgery to the bone; and only surgery to the bone.

Outcome Measures
The primary study outcomes were payer costs and length of stay (LOS). Payer cost per SRE-associated hospitalization was measured as reimbursed amount paid by the health plan to a hospital, including the Medicare-paid portion, after subtracting patient copayments and deductibles. SRE-associated hospital costs included reimbursement of all claims for the full hospital stay until discharge for SRE-related hospitalizations as defined above. Start and end dates of service using room and board revenue codes were identified and all services within that duration, including professional services, were attributed to the hospital stay. Emergency room visits that resulted in the direct admission to the hospital were captured as inpatient admission costs. All costs were adjusted to June 2009 dollars using the medical care component of the Consumer Price Index. 22 LOS per SRE-associated hospitalization was calculated and was assessed by type of cancer and type of SRE.

Statistical Analysis
Payer costs and LOS for SRE-associated hospitalizations were examined using descriptive analyses that included descriptive profiles of patients' demographic characteristics and hospital discharge status. In bivariate analyses, the distributions of payer cost and LOS per hospitalization were examined by type of cancer and type of SRE. In anticipation of a skewed cost distribution, we planned a priori to conduct a sensitivity analysis excluding extremely low or high costs of hospitalizations within the upper 1% and lower 1% of total costs to estimate their impact on mean payer costs and LOS.
To handle uncertainties in possible small sample sizes for less frequently occurring SREs (e.g., spinal cord compression), CIs for mean costs were estimated by the bootstrap method (repeated resampling with replacement from the original sample). 23 The benefit of performing bootstrapping is that only one assumption is required, which is that the patients in this study were representative of the typical patient with advanced prostate or breast cancer or myeloma experiencing these events. We randomly selected a new sample of patients (with repeat patient selection possible) for each cohort of cancer and SRE type and estimated mean costs for each. We repeated this sampling 1,000 times to generate 1,000 estimates of mean cost for each cohort of SRE by tumor type. Confidence intervals (95%) were then constructed based on these 1,000 randomly drawn samples. All descriptive analyses were conducted using SAS version 9.1 (SAS Institute Inc., Cary, NC), and bootstrapping was conducted using Stata 11 (StataCorp, College Station, TX).

■■ Results Demographic Characteristics and Discharge Status
Patient demographic characteristics are reported in Table 1. The mean (SD) ages in years were 72.1 (11.6) for patients with  cancer types, 58%-77% of patients were discharged to home, 11%-18% to rehabilitation facilities, and 6%-8% to acute care facilities. Between 1% and 3% of patients died during an SREassociated hospitalization. A total of 572 hospitalizations for prostate cancer, 1,542 for breast cancer, and 1,495 for multiple myeloma met the inclusion criteria (Figure 1). Between 4.5% and 6.4% (n = 217) of hospitalizations were excluded from the study across the cancer types because they occurred within 30 days of a previous SRE hospitalization discharge date. For all 3 types of cancer, using prostate cancer, 61.3 (12.8) for patients with breast cancer, and 65.4 (12.5) for patients with multiple myeloma. Approximately 70% of patients with prostate cancer, 33% of patients with breast cancer, and 47% of patients with multiple myeloma were 65 years or older. Across all cancer types, between 65% and 73% of patients lived in the southern or north central areas of the United States. This distribution reflects the regional distribution of covered lives in the MarketScan database and does not necessarily reflect the regional distribution of SRE-associated admissions in the United States. Across the  bone, $22,390 ($28,042) for pathologic fracture, and $59,788 ($66,466) for spinal cord compression. Mean (SD) LOS per admission was 7.7 (9.1) days for surgery to the bone, 6.6 (5.6) days for pathologic fracture, and 11.6 (10.4) days for spinal cord compressions. For patients with breast cancer, mean (SD) payer costs were $32,742 ($34,836) for surgery to the bone, $26,936 ($29,727) for pathologic fracture, and $59,854 ($59,334) for spinal cord compression. Mean (SD) LOS was 6.1 (5.3) days for surgery to the bone, 6.8 (6.2) days for pathologic fracture, and 9.4 (8.2) days for spinal cord compressions.
After excluding the upper 1% and lower 1% of reimbursement costs, mean cost estimates declined across all tumor types, ranging from $29,516 to $36,165 for surgery to the bone, $20,855 to $25,447 for pathologic fracture, and $39,957 to $58,012 for spinal cord compression. For prostate cancer, mean cost estimates decreased by 14.1% for patients experiencing surgery to the bone, 6.9% for pathologic fracture, and 6.5% for spinal cord compression. For breast cancer, mean cost estimates decreased by 6.1% for surgery to bone, 5.5% for the mutually exclusive classification method, hospitalization occurred most often for pathologic fracture with or without surgery to bone ( Figure 2, Table 2). For prostate cancer, 21.3% of hospitalizations were categorized as surgery to the bone, 69.9% as pathologic fracture (with or without surgery to the bone), and 8.7% as spinal cord compression (with or without pathologic fracture and/or surgery to the bone). In the breast cancer cohort, 15.0% of hospitalizations were categorized as surgery to the bone, 79.0% as pathologic fracture, and 6.0% as spinal cord compression. In the multiple myeloma cohort, 23.6% of hospitalizations were categorized as surgery to the bone, 68.3% as pathologic fracture, and 8.1% as spinal cord compression.
Diagnosis and procedure codes for more than 1 SRE type were present in 244 (42.7%) of hospitalizations in the prostate cancer group, in 760 (49.3%) for breast cancer, and in 518 (34.6%) for multiple myeloma. The proportions of SRE code combinations that occurred under the spinal cord compression and pathologic fracture categories were generally similar across tumor types (Figure 2).  pathologic fracture, and 3.1% for spinal cord compression. For multiple myeloma, mean cost estimates decreased by 4.8% for surgery to bone, 8.0% for pathologic fracture, and 8.5% for spinal cord compression. Table 4 presents the bootstrapped 95% CIs of mean payer costs per hospitalization (based on all hospitalizations including those with reimbursed costs in the upper 1% and lower 1%).

■■ Discussion
Bone lesions secondary to advanced malignancies are a common occurrence and can have devastating clinical consequences in patients. Patients with bone lesions can experience burdensome SREs such as pathologic fracture, spinal cord compression, or need for surgery or radiation to bone. These skeletal complications are also important from a health economic perspective in prostate and breast cancers, given the prevalence of these diseases, and in multiple myeloma due to the extent of bone destruction that occurs with the disease.
In this study, we used 2 large national claims databases to estimate the mean costs that third-party payers incur for reimbursing SRE-associated hospitalizations (excluding radiation to bone, hypercalcemia of malignancy, and bone pain) among patients with multiple myeloma or bone metastasis secondary to prostate or breast cancer. Mean hospital reimbursements were higher across tumor types for spinal cord compression with or without pathologic fracture and/or surgery to the bone ($43,691-$59,854) than for surgery to the bone only ($31,016-$42,094) or pathologic fracture with or without surgery to the bone ($22,390-$26,936).
The mutually exclusive categorizations used to classify claims under 1 SRE when more than 1 SRE code was present allowed us to develop a composite view of each SRE, thus better reflecting what is occurring in patients. For example, the composition of spinal cord compression as an SRE allowed up to 4 possible combinations of coding: spinal cord compression only, spinal cord compression plus pathologic fracture, spinal cord compression plus surgery to the bone, or spinal cord compression plus pathologic fracture and surgery to bone. Since each of these 4 possibilities is represented in the spinal cord compression claims selected in our study across tumor types, we are more confident that the average costs calculated here provide a reliable estimate of the inpatient costs.
Lastly, we conducted sensitivity analyses to account for skewed cost distributions and potentially small samples of patients in the subgroups of SRE and cancer type. After removing the outliers, the mean cost estimates did not change substantially for the majority of SREs across tumor types.

Limitations
First, the study was limited to SREs in inpatient settings only. Results do not represent the cost of SREs overall or of treatments provided in outpatient settings. Moreover, the study patients represent only a portion of patients with SREs because not all patients with SREs are treated in inpatient settings. Second, the results represent net payer cost after subtracting patient copayments and deductibles, rather than total allowed cost. To address the possibility that exclusion of patient paid amounts affected our findings, we analyzed the percentage of costs paid by patients and found that it was small, ranging from 1.5% to 4.2% of total cost depending on tumor type and SRE type. Third, a common limitation in studies using administrative health care databases is selecting claims based on ICD-9-CM and CPT codes that accurately represent the specific population of interest and clinical endpoints such as SRE-associated hospitalizations. To minimize the potential bias in selecting claims and improve the specificity for selecting the correct population in our study, we consulted with an independent coding expert to identify a list of codes for extracting SRE-associated hospitalizations for this analysis. Fourth, our analysis is based on a sample of patients covered by health care plans from medium to large employers and some who are also covered by Medicare (dual-eligibility). The results from our study may not reflect the full spectrum of reimbursed amounts for these SRE-associated hospitalizations and might have limited generalizability. Fifth, we assumed that patients identified in this database are typical patients with cancer and bone metastasis. Patient age in these populations is consistent with the SEER database; 24 however, no further comparisons were made.

■■ Conclusions
The inpatient costs associated with treating SREs are significant from a payer perspective. The present study used a careful process for patient selection and a mutually exclusive categorization method that assessed heterogeneity among SRE types and tumor types. This study provides an estimate that is helpful to payers in quantifying the inpatient financial impact of cancer-related SREs. SREs place a substantial burden on patients with advanced tumors and on third-party payers.