A retrospective analysis of health care costs for bone fractures in women with early-stage breast carcinoma

Authors

  • Zhiyuang Zhou Ph.D.,

    1. Global Outcomes Research, Pharmacia Corporation, Bridgewater, New Jersey
    Search for more papers by this author
    • Dr. Zhou is an employee of Pfizer Inc.

  • Alberto Redaelli Ph.D.,

    Corresponding author
    1. Global Outcomes Research Oncology, Pharmacia Corporation, Milan, Italy
    • Global Outcomes Research-Oncology, Pharmacia Corporation, Via Robert Koch 1.2 (20152) Milan, Italy
    Search for more papers by this author
    • Fax: (011) 39 0248382392

    • Dr. Redaelli is an employee of Pfizer Inc., and owns stock options in both Pfizer and Pharmacia.

  • Olof Johnell M.D., Ph.D.,

    1. Department of Orthopedics, University Malmó General Hospital, Malmó, Sweden
    Search for more papers by this author
  • Richard J. Willke Ph.D.,

    1. Global Outcomes Research, Pharmacia Corporation, Peapack, New Jersey
    Search for more papers by this author
    • Dr. Willke is an employee of Pfizer Inc., and owns both Pfizer stock and stock options.

  • Giorgio Massimini M.D.

    1. Medical Development, Pharmacia Corporation, Nerviano, Milan, Italy
    Search for more papers by this author
    • Dr. Massimini is employed by Pfizer Inc. and owns Pfizer stock.


Abstract

BACKGROUND

In this retrospective data base study, the authors sought to estimate direct costs for bone fractures in women age ≥ 65 years with early-stage breast carcinoma and to compare those costs with treatment costs for bone fractures in older women without early-stage breast carcinoma.

METHODS

Direct costs for bone fractures in patients with early-stage breast carcinoma, which consist of excess treatment costs for bone fracture and excess costs of long-term care for bone fracture, were evaluated by using the 1997–1998 Standard Analytical File. The statistical significance of the difference in inpatient costs, medical treatment costs, and long-term care admission rates were determined with the t test and the Fisher chi-square test, respectively.

RESULTS

For older women with early-stage breast carcinoma, the direct costs for bone fracture were estimated at $45,579, and 57% of those costs came from treating the bone fracture (32% came from inpatient hospital costs, and 25% came from noninpatient hospital costs), 25% came from other excess treatment costs, and 18% came from excess long-term care costs. The women who had early-stage breast carcinoma and sustained bone fracture did not differ significantly from the women without early-stage breast carcinoma who sustained a bone fracture.

CONCLUSIONS

Bone fracture was associated with high direct costs in older women with early-stage breast carcinoma. Additional research should include appropriate, incidence-based studies to investigate the potential benefit of an intervention for preventing bone fracture in this increasingly large patient population. Cancer 2004. © 2003 American Cancer Society.

In western countries, the breast is the most common site of cancer in women. Bone is the prevalent site of metastasis in patients with breast carcinoma. The spread of breast carcinoma to bone can cause significant pain and decreased activity. Pathologic fractures can be devastating because of decreased mobility from femoral fractures or, in some patients, spinal cord compromise with vertebral body collapse. Given the current prevalence of metastatic breast cancer in women in the U.S. (exceeding 100,000 diagnoses per year), measures to reduce the morbidity from skeletal involvement by breast carcinoma in the early stages are important for maintaining a patient's quality of life and for controlling the substantial financial consequences of the metastatic bony involvement. In fact, Burge et al.1 recently showed that approximately 700,000 vertebral fractures occur annually in men and women in the U.S., at an estimated cost of $746 million in 1997. Women with early-stage breast carcinoma who are treated with antiestrogens, and especially the new aromatase inhibitors and inactivators, over a long period (for example, 5–10 years) may face an increased risk of bone fractures. If this hypothesis is confirmed, then the medical concerns would be compounded by the economic burden from the costs of a higher incidence and prevalence of bone fractures compared with the normal, age-matched, female population. Development of a model to measure the economic extent of this problem requires reliable prevalence and cost data. However, to our knowledge, the literature provides few economic data with which to create such a model. Therefore, we conducted a retrospective data base study to provide this information. Our objectives were to estimate the health care costs for bone fractures in women age ≥ 65 years who had early-stage breast carcinoma and to compare those costs with the costs for bone fractures incurred by older women who did not have breast carcinoma.

MATERIALS AND METHODS

Data Sources

The data base that was used for this study was the 1997–1998 Standard Analytical File (SAF) (Medicare 5% national sample), which is representative of the Medicare population in the U.S. The SAF data base records medical claims for inpatient hospital, outpatient hospital, skilled-nursing facility, home health care, hospice, and professional services. It is a reliable and valuable source for the estimation of costs for bone fractures. These 2-year data, however, although they identify diagnoses of breast carcinoma, do not provide cancer staging information, which would have helped us identify patients with early-stage breast carcinoma. To overcome this obstacle, we developed an algorithm to exclude women with symptoms of late-stage breast carcinoma.

Study Cohorts

Early-stage breast carcinoma

The cohort of patients with early-stage breast carcinoma included women age ≥ 65 years on January 1, 1997 who had been enrolled in Medicare Parts A or B for 2 full years (1997 and 1998), had not been enrolled in a Medicare managed-care program, and had diagnoses defined by International Classification of Diseases Ninth Revision (ICD-9) codes for malignant neoplasm of the female breast (174.XX). Because the SAF provides no cancer staging information, obtaining data only on women with early-stage breast carcinoma was a major challenge. To increase the likelihood that the cohort of patients with breast carcinoma who were included in this study had early-stage breast carcinoma, we excluded women who met the following criteria for advanced metastatic neoplasm:

1) Patients were excluded if they had additional cancer diagnoses (ICD-9 codes: malignant neoplasm, 140.XX–173.XX and 175.XX–208.XX). This criterion excluded patients with metastatic cancer who had secondary neoplasm codes, signaling advanced breast carcinoma. The criterion also ensured that breast carcinoma in this cohort was primary cancer.

2) Patients who subsequently died were excluded from the study. Death due to cancer would have been another sign of advanced metastatic disease. Although we would have preferred to exclude from the cohort only women who died of cancer, the SAF contained no information regarding the cause of death other than inpatient hospital discharge status. Because few patients died in the hospital, we excluded all patients who died during the study period (1997–1998).

3) Patients who subsequently developed malignant pleural effusions also were excluded. This condition develops at some time in nearly 50% of patients with metastatic breast carcinoma2; therefore, it would be a specific identifier for advanced disease.

4) Finally, patients who developed recurrent breast carcinoma were excluded from the study. The method of identification was similar to that used by Earle et al.3 for acute myelogenous leukemia, with the exception of the length of disease remission (no treatment) period. We considered patients who received a second treatment after 4 months of remission to have developed a recurrence.

Early-stage breast carcinoma with bone fracture

To determine the cohort of patients with early-stage breast carcinoma who sustained bone fracture, we collected data concerning all bone fractures (ICD 9 codes: bone fractures, 800XX–829XX; pathologic fractures, 7331) within the early-stage breast carcinoma cohort. To increase the likelihood that bone fractures in these patients were related causally to osteoporosis and breast carcinoma, individuals who had accident codes or E-codes on their medical record (indicating severe trauma) as a secondary diagnosis were excluded. We also excluded women who had only ICD-9 codes for open fracture, which may have been caused by factors other than deterioration of osteoporosis. Women who had at least 1 diagnosis of bone fracture within the 2-year period were identified with bone fractures and were categorized further into two groups—those who had ever been hospitalized for bone fracture and those who had never been hospitalized for bone fracture—based on a review of their inpatient hospital claims for 2 years.

Age-matched and comorbidity-matched subcohort of women with early-stage breast carcinoma and no bone fracture

We used stratified, random sampling to create two bone-fracture subcohorts from the cohort of women with early-stage breast carcinoma without bone fractures: one matched for age and comorbidity to the ever-hospitalized group and the other matched to the never-hospitalized group. We stratified each bone fracture subcohort into 15 combined categories, combining 5 age categories (ages 65–69 years, 70–74 years, 75–79 years, 80–84 years, and ≥ 85 years) with 3 comorbidity groups (0–5 comorbid conditions, 6–10 comorbid conditions, and ≥ 11 comorbid conditions). The age-matched, normal cohort was used to compare the costs of bone fracture in patients who had early-stage breast carcinoma with the costs of bone fracture in normal women, thus estimating the excess costs for bone fracture and excess long-term care admission rates.3 The women from the normal population met the following criteria: 2 full years (1997 and 1998) of Medicare enrollment (Part A and Part B), no enrollment in a Medicare managed-care program, and no diagnoses defined by ICD-9 codes for malignant neoplasm (140.XX–208.XX) in 1997 or 1998. In addition, women in the normal cohort met the same criteria that were used for the early-stage breast carcinoma cohort to identify women with bone fractures. This cohort also was separated into two groups, ever or never hospitalized for bone fracture, based on review of their inpatient hospital claims for the 2 years.

Cost Analyses

The Medicare program for American seniors (Parts A and B) covers most medical care costs (except for long-term care and outpatient drugs) and records all charges and payments for the medical care it covers. Using the Medicare 5% sample data (1997–1998), we evaluated direct costs for bone fractures in patients with early-stage breast carcinoma, which included the estimation of medical treatment costs for bone fracture, excess treatment costs for bone fracture, and excess long-term care costs for bone fracture. Furthermore, we compared the inpatient costs, medical treatment costs, and long-term care admission rates between the cohort of women with bone fracture and early-stage breast carcinoma and the cohort of women with bone fracture and without early-stage breast carcinoma, stratified by age group, status of hospitalization for bone fracture, and type of bone fracture. The statistical significance of the difference in means of costs and long-term care admission rates was tested with the Student t test and the Fisher chi-square test, respectively.

Definitions: Components of Direct Costs for Bone Fracture

Medical treatment costs for bone fracture were defined as the total of all treatment costs for bone fracture. In practice, all medical claims with ICD-9 codes of bone fractures for patients in the early-stage breast cancer cohort were identified. For institutional claims (inpatient hospital, outpatient hospital, home health, and skilled-nursing facility), only claims with primary diagnosis codes of bone fractures were included. First, we estimated the resource-utilization and direct costs for each cohort. Our overall strategy for estimating direct costs was to use cost-to-charge ratios for institutional claims and to use Medicare-allowed charges as a proxy for costs for ambulatory care services. If allowed charges were not available, then we used total payments (insurance reimbursements plus patients' out-of-pocket expenditures) as our measure of cost.

Excess treatment costs for bone fracture were defined as the difference in total treatment costs between the early-stage breast carcinoma with bone fracture cohort and the early-stage breast carcinoma without bone fracture cohort, assuming that other characteristics of the two subcohorts were equal. The rationale for this definition was that bone fracture may have hidden economic consequences in addition to direct treatment costs, such as inpatient hospitalization with bone fracture as a secondary diagnosis and the treatment necessitated by the deterioration of physical condition due to bone fracture. Both subcohorts were matched for age and comorbidity to ensure a fair comparison. Then, we summed costs for all medical claims, regardless of the reasons for treatment, within each subcohort.

Excess long-term care admission rate for bone fracture was defined as the difference in long-term care admission rates between the two early-stage breast carcinoma cohorts, again assuming that other characteristics of the two subcohorts were equal. Because Medicare does not cover long-term care costs, a direct record was not available. In addition, SAF data did not include information regarding the discharge disposition of hospital inpatients, which one author used to identify long-term care admissions.4 Our strategy, however, was to use the item place of service (code 31, skilled nursing facility; code 32, nursing facility; code 33, custodial care facility; and code 35, adult living care facility) in the professional service file (Part B file) to identify patients staying in long-term care facilities.

Excess long-term care costs for bone fracture were defined as the product of long-term costs per individual admitted for bone fracture and the excess long-term care admission rate for bone fracture. To our knowledge, the only study that evaluated the long-term cost per individual admitted for bone fracture was the Office of Technology Assessment (OTA) study, which estimated the index for hip fracture in a population of individuals age ≥ 50 years in 1990.5 We adjusted the average per-person expenditure for patients with hip fracture who were discharged from the hospital to a nursing home ($13,849 for 1990; adapted from the OTA study) and the excess admission rate for bone fracture to 1997 price levels by multiplying it by the ratio of medical price indices for 1996 to 1991.

Direct costs for bone fracture were the sum of excess treatment costs for bone fracture and excess long-term costs for bone fracture, as defined above.

RESULTS

Prevalence of Bone Fractures in the Early-Stage Breast Carcinoma Cohort, 1997–1998

Figures 1 and 2 show the rates of bone fractures in patients with early-stage breast carcinoma age ≥ 65 years by age group, type of fracture, and status of hospitalization for 1997–1998 (data on file). Overall, 2486 of 20,887 women (11.9%) in the early-stage breast carcinoma cohort had at least 1 type of fracture. Among patients with fractures, approximately 20% were ever hospitalized for any fracture during the 2-year study period. The prevalence of any hospitalization for any fracture increased from 1.0% for women ages 65–69 years to 7.3% for women age ≥ 85 years, an increase of approximately 7-fold; in contrast, the prevalence of no hospitalization for any fracture increased from 7.3% to 13.0%, almost doubling. The prevalence for fracture of the hands and feet (31.6 per 1000 patients) was the highest, followed by the prevalence for fracture of the clavicle, humerus, radius, and ulna (29.2); fracture of the femoral neck (24.0); fracture of the ribs (19.2); and fracture of the vertebral column (17.3). The prevalence of any hospitalization for bone fracture was substantially lower compared with the prevalence of no hospitalization for bone fracture for all types of fractures, with the exception of fracture of the femoral neck (13.4 per 100 patients vs. 10.6 per 1000 patients). Women who were ever hospitalized probably had severe or new fractures, whereas women who were never hospitalized may have included moderate bone fractures or older patients who were hospitalized before the study period.

Figure 1.

Rate of bone fractures in patients with early-stage breast carcinoma (EBC) age ≥ 65 years by age and status of hospitalization, 1997–1998.

Figure 2.

Rate of bone fractures in patients with early-stage breast carcinoma (EBC) age ≥ 65 years stratified by type of fracture and status of hospitalization, 1997–1998.

Hospital Admission Rate, Length of Stay, and Inpatient Costs per Person for Bone Fractures

Table 1 provides the mean costs for women in the early-stage breast carcinoma with bone fracture cohort who were ever hospitalized during the 2-year period (1997–1998) according to patient age and type of bone fracture. Hospital admission, length of stay, hospital admission rate, and inpatient costs per person are also reported. The mean cost for hospital admission was $12,498 (standard deviation [SD], $12,042), the mean length of stay was 5.7 days (SD, 4.5 days), and the mean cost per person for inpatient hospital services within the 2-year period was $14,604 (SD, $13,756). Approximately 17% of patients in the ever-hospitalized group were readmitted to the hospital. The mean inpatient costs per person were highest for women ages 65–69 years ($19,866), dropped to $11,854 for women ages 70–74 years, and increased to $15,800 for women ages 80–84 years. The highest cost per person for the women ages 65–69 years was affected largely by a single outlier. Examining the inpatient costs per person by type of fracture, we found that fracture of the femoral neck/other part of femur was the most expensive category ($18,058); followed by fractures of the patella, tibia and fibula, and ankle ($13,399); and pathologic fractures ($9963). The costs for the patients with other fractures ranged from $7500 to $8500. Several possible reasons may explain the high standard deviation in this table and some of the other tables: Within a category of fracture, we grouped many codes. We do not have severity indicators for the development of homogeneous fracture categories, and the numbers of fractures were relatively small.

Table 1. Hospital Admission Rate, Mean Hospital Stay, and Cost per Person for Bone Fractures Stratified by Age Group and Type of Fracture in Ever-Hospitalized Patients with Early Breast Carcinoma
VariableNo. of bone fractures in cohortHospital admission rate (%)aCost per admission ($)Length of stay per admission (days)Cost per person ($)
MeanSDMeanSDMeanSD
  • SD: standard deviation.

  • a

    Admission rates were > 100%, because all included patients were in the “ever-hospitalized” cohort.

  • b

    Codes in parentheses are from the International Classification of Diseases, 9th revision.

Age group (yrs)        
 65–6944111.417,83931,6246.68.719,86633,236
 70–74113111.510,63170055.13.711,8547874
 75–79153117.612,11291925.94.714,24911,776
 80–84160121.912,96488135.83.715,80011,092
 ≥ 8564115.611,84669455.52.713,6977355
 Subtotal534116.912,49812,0055.74.514,60413,756
Type of bone fractureb        
 Pathologic fracture (7331)62111.3895278956.26.299638207
 Fracture of vertebral column with or without spinal cord injury (805, 806)43114.0747772286.04.185208408
 Fracture of rib(s), sternum, larynx, and trachea/pelvis (807, 808)53107.5695464085.03.274796741
 Fracture of clavicle, humerus, radius/ulna (810, 812, 813)54101.9846265883.52.986186783
 Fracture of hands/feet (814, 815, 816, 817, 825, 826)5100.0556238643.61.555623864
 Fracture of femoral neck/other part of femur (820, 821)280111.416,20610,8696.44.618,05811,990
 Fracture of patella, tibia/fibula, ankle (822, 823, 834)63114.311,72421.2224.83.813,39923,350
 Other fractures (809, 811, 818, 819, 827, 829)3166.7486912495.42.781167742

Medical Treatment Costs per Person

Table 2 presents the substantial differences in medical treatment costs per person (inpatient hospital, outpatient hospital, home health, skilled-nursing facility, and other charges) between patients in the ever-hospitalized group and the never-hospitalized group among the cohort of women with early-stage breast carcinoma with bone fracture stratified by status of hospitalization, age group, and type of bone fracture. Patients in the ever-hospitalized group had costs of $26,033 (SD, $22,307), whereas patients in the never-hospitalized group had costs of only $1385 (SD, $4333). Inpatient cost was the most important component of the medical treatment costs to Medicare of bone fractures (56.1%) for the ever-hospitalized group, followed by the costs for skilled nursing facilities (22.7%) and costs for professional services (13.9%). Moreover, the ratio of medical treatment costs to inpatient hospital costs increased by age, from 1.43 for women ages 65–69 years to 1.94 for women age ≥ 85 years. The distribution of average costs for bone fractures by age group was bimodal; the average cost of $28,533 for the group ages 65–69 years was a high, and the average cost of $30,461 for the group ages 80–84 years was another high (Fig. 3). However, as we have indicated, a single outlier affected the highest value of average treatment costs for women ages 65–69 years. Table 3 shows that, for those ever hospitalized for a specific type of fracture, fracture of the femur was the most costly ($31,809; SD, $18,727), followed by fractures of patella, tibia and fibula, and ankle ($22,004; SD, $25,709). The average treatment costs for other types of fracture were much lower, ranging from $11,000 to $15,000.

Table 2. Inpatient Hospital, Outpatient Hospital, Home Health Care, Skilled Nursing Facility, and Other Costs for Fractures Stratified by Age Group in Patients with Breast Carcinoma
Age group (yrs)Cohort size (no. of patients)Mean cost per person ($)
Inpatient hospitalOutpatient hospitalHome health careSkilled nursing facilityOthersTotal
MeanSD
  1. SD: standard deviation.

Ever hospitalized for specific fracture        
 65–694419,86654127342734421328,53341,682
 70–7411311,85434712283301321519,94514,091
 75–7915314,24948013785240344724,79420,121
 80–8416015,80045316458680388230,46123,253
 ≥ 856413,69721319197404372326,95714,292
 Subtotal53414,60441714755913362526,03322,307
Never hospitalized for any fracture        
 65–693184461191748110622338
 70–7453440811933349513556248
 75–7955135027228051314163419
 80–8443239734430249315363303
 ≥ 8511729531869738416945006
 Subtotal195238822428249113854333
Figure 3.

Average costs for treatment of bone fractures for those ever hospitalized for bone fractures stratified by age, 1997–1998.

Table 3. Inpatient Hospital, Outpatient Hospital, Home Health Care, Skilled Nursing Facility, and Other Costs for Fractures Stratified by Hospitalization Status and Type of Bone Fractures in Patients with Breast Carcinoma
VariableaCohort size (no. of patients)Mean cost per person ($)
Inpatient hospitalOutpatient hospitalHome health careSkilled nursing facilityOthersTotal
MeanSD
  • SD: standard deviation.

  • a

    Codes in parentheses are from the International Classification of Diseases, 9th revision.

Ever hospitalized for specific fracture        
 Pathologic fracture (7331)62996334446239418913,02611,690
 Fracture of vertebral column with or without spinal cord injury (805, 806)4399631741440327992914,34214,207
 Fracture of rib(s), sternum, larynx, and trachea/pelvis (807, 808)53852061716392688113,0629374
 Fracture of clavicle, humerus, radius/ulna (810, 812, 813)5474794997552502289015,26413,712
 Fracture of hands/feet (814, 815, 816, 817, 825, 826)58618165732406877411,3028917
 Fracture of femoral neck/other part of femur (820, 821)28018,05835514667378455131,80918,727
 Fracture of patella, tibia/fibula, ankle (822, 823, 834)6313,39942710534298282722,00425,709
 Other fractures (809, 811, 818, 819, 827, 828, 829)38116110810693306715
Never hospitalized for specific fracture        
 Pathologic fracture (7331)2711252942732078992126
 Fracture of vertebral column with or without spinal cord injury (805, 806)3181822101293238432073
 Fracture of rib(s), sternum, larynx, and trachea/pelvis (807, 808)3491632262161697732385
 Fracture of clavicle, humerus, radius/ulna (810, 812, 813)55670525011980918823088
 Fracture of hands/feet (814, 815, 816, 825, 826)65619174372635651252
 Fracture of femoral neck/other part of femur (820, 821)2221603481621819294812,125
 Fracture of patella, tibia/fibula, ankle (822, 823, 834)28132017623448412143062
 Other fractures (809, 811, 818, 819, 827, 828, 829)149151221791686181582

Excess Costs, Excess Long-Term Care Admission Rates, and Direct Costs of Bone Fracture

Table 4 shows that the excess costs, as defined previously, were $37,179 for women who were ever hospitalized for bone fracture and $4722 for women who were never hospitalized for bone fracture. The difference of the excess costs between age groups was moderate for the first group and large for the second group. Table 5 shows that the excess long-term care admission rate was 44.8% for women who were ever hospitalized for bone fracture and 5.5% for women who were never hospitalized. In the ever-hospitalized cohort, the excess admission rate increased from 25.0% for women ages 65–69 years to 40.7% for women ages 70–74 years and remained stable for the other age groups. For the never-hospitalized cohort, the excess admission rate increased with the age of the patients with early-stage breast carcinoma and bone fracture. The direct costs for bone fracture for the ever-hospitalized cohort accounted for a total of $45,579, which was comprised of excess medical treatment costs for bone fracture ($37,179) and excess long-term costs for bone fracture ($8400). The excess medical treatment costs for bone fracture could be factored further into two components: $26,033 for the treatment cost of bone fracture and $11,146 for other excess costs of bone fracture. Women who were never hospitalized for bone fracture, as expected, had much smaller direct costs for bone fracture ($5783).

Table 4. Excess Costs for Bone Fractures Stratified by Age Group in Patients with Breast Carcinoma
Age group (yrs)EBC cohort  
Bone fracture from all causesNonbone fracture from all causesExcess costs
Mean ($)SD ($)No.Mean ($)SD ($)No.Mean ($)P value
  1. EBC: early breast carcinoma; SD: standard deviation.

Ever hospitalized for specific fracture        
 65–6963,85469,5904429,91638,2594438,1680.0022
 70–7416,85488,17811330,24153,4461333,6130.0007
 75–7971,46369,63915333,07743,56815338,387<0.0001
 80–8470,59147,19716032,65640,26716037,935<0.0001
 ≥ 8569,13052,1986431,11247,6946438,017<0.0001
 Subtotal69,03366,35053431,85544,87553437,179<0.0001
Never hospitalized for any fracture        
 65–6927,14168,05231818,16026,76232389810.0289
 70–7426,39036,95753421,52731,00954748630.0194
 75–7930,22442,81555125,90636,47955543170.0714
 80–8431,06737,45043229,92148,51843711460.6957
 ≥ 8534,50347,32011727,21333,92112173990.1683
 Subtotal29,11645,727195224,39436,839198347220.0004
Table 5. Excess Long-Term Care Admission Rates for Bone Fractures Stratified by Age Group in Patients with Early Breast Carcinoma
Age group (yrs)EBC cohort  
Bone fractureNonbone fracture  
No. of patientsAdmission (%)No. of patientsAdmission (%)Excess admission rate for fracture
Fractures in cohortAdmission to to LTCRateSDFractures in cohortAdmission to to LTCRateSDRate (%)P value
  1. EBC: early breast carcinoma; LTC: long-term care; SD: standard deviation.

Ever hospitalized for specific fracture          
 65–69441329.56.94424.53.125.00.0018
 70–741335346.94.711376.32.340.7<0.0001
 75–791539360.83.91532013.12.747.7<0.0001
 80–8416011068.83.71603220.03.248.8<0.0001
 ≥ 85645179.75.0642031.35.848.4<0.0001
 Subtotal53432059.92.15348115.21.644.8<0.0001
Never hospitalized for any fracture          
 65–69318237.21.532392.80.94.40.0098
 70–74534489.01.2547305.51.03.50.0260
 75–795518315.11.55555610.11.35.00.0126
 80–8443211225.92.14378319.01.96.90.0143
 ≥ 851174639.34.51212924.03.915.30.0108
 Subtotal195231216.00.8198320710.40.75.5<0.0001

Comparison of the Early-Stage Breast Carcinoma Cohort with the Normal Cohort

Table 6 shows that the medical treatment costs of bone fracture per person did not differ significantly between the cohorts of women with early-stage breast carcinoma and bone fracture and normal women with bone fracture, either stratified by status of hospitalization for bone fracture or stratified further by age group. In a supplemental study (data on file), we also found no significant difference in medical treatment costs for bone fracture per person between these two cohorts stratified by status of hospitalization and type of bone fractures (data on file). Long-term care admission rates also were compared between the cohorts of women with early-stage breast carcinoma and bone fracture and normal women with bone fracture (data on file). For the women who were ever hospitalized for bone fracture, the long-term care admission rate was slightly higher in the cohort that included women with early-stage breast carcinoma and bone fracture, but the difference was not statistically significant (P = 0.460). Conversely, for the women who were never hospitalized for bone fracture, the long-term care admission rate was significantly lower (5.1%) in the cohort that included women with early-stage breast carcinoma and bone fracture (P < 0.0001). It is not clear why the long-term care admission rate was lower in this cohort: it is possible that, compared with the cohort of normal women, these women may have been affected more by uncomplicated bone fractures or may have stayed at home more frequently, where more family or proxy support was available for their compromised health status.

Table 6. Differences in Medical Treatment Costs per Person between Bone Fractures in Women from the Early Breast Carcinoma Cohort and from the Cohort of Normal Women Based on Status of Hospitalization and Age Group
Age group (yrs)Treatment cost for bone fractures  
EBC cohortNormal womenDifference
Mean ($)SD ($)No.Mean ($)SD ($)No.Mean ($)P value
  1. EBC: early breast carcinoma; SD: standard deviation.

Ever hospitalized for specific fracture        
 65–6928,53341,6824422,30419,9754362290.38
 70–7419,94514,09111320,38912,96889−4440.82
 75–7924,79420,12115323,30017,98215714940.49
 80–8430,46123,25316025,80116,38916146590.04
 ≥ 8526,95714,2926428,86816,77663−19120.49
 Subtotal26,03322,30753424,18016,87451318530.13
Never hospitalized for any fracture        
 65–691062233831811842324273−1220.53
 70–741355624853411342549466−2210.45
 75–791416341955115423609515−1270.56
 80–841536330343216694716422−1330.63
 ≥ 851694500611718594859117−1650.80
 Subtotal138543331952143236141793−480.72

DISCUSSION

This study provides new evidence that bone fractures in women with early-stage breast carcinoma are costly to the medical care system, especially for patients who require hospitalization for the bone fracture. In older women with early-stage breast carcinoma, the direct cost for bone fracture was estimated at approximately $45,579, 57% of which comes from treating the fracture itself (32% from inpatient hospital costs and 25% from noninpatient hospital costs), 25% of which comes from other excess treatment costs, and 18% of which comes from excess long-term care costs (Fig. 4). To our knowledge, this is the first study in which an attempt was made to estimate the excess treatment costs and excess long-term care costs associated with bone fractures in a population of older women with early-stage breast carcinoma. We found that the cost estimates for bone fractures reported in this study were consistent with the cost estimates published in the literature.4–8 For comparison, we excluded other excess treatment costs ($11,146), which previous studies did not account for, from direct costs for bone fractures ($45,579). The new estimate of $34,433 represents approximately 1 year of bone fracture costs at 1997 prices. This value must be compared with the values shown in the previously cited studies (which used 1995 per-patient costs for hip fracture): $27,318,5 $32,428,4 $34,696,6 and $35,104.7 Even adjusting for some difference in methods, characteristics of the selected women cohorts, and types of fractures, our estimated per-patient costs for bone fracture for women who were ever hospitalized for bone fracture were notably close to the estimates from those previous studies. Inpatient hospital costs, medical treatment costs, and long-term care admission rates did not differ significantly between the group of women who were ever hospitalized for bone fracture and the group of women who were never hospitalized for bone fracture, with one exception: for never-hospitalized women, the long-term care admission rate was significantly lower in the early-stage breast cancer cohort.

Figure 4.

Average direct costs for bone fractures for those ever hospitalized for bone fractures stratified by type of service, 1997–1998.

Distinct Features of This Study

In addition to our focus on bone fractures in women with early-stage breast carcinoma, our study had three distinctive features: First, we applied a “bottom-up” approach that estimated the cost per person directly based on a single, individual-level data base; whereas most previous studies on the costs of bone fractures4–7, 9, 10 were national, prevalence-based, cost-of-illness studies that gathered data from several sources (mostly surveys) and generated consensus costs estimates but did not provide a framework for assessing costs on a per-patient basis.4 One study8 used hospital discharge data from a nationwide inpatient sample to estimate the inpatient hospital costs per admission. Although the inpatient hospital cost is a major component of the costs for bone fracture, we believe that missing information from the other cost components prevents the reader from calculating the full economic consequences of bone fracture. The second unique feature of our study is its estimation of the excess treatment costs and excess long-term costs for bone fracture. In our opinion, estimating excess cost has two advantages: First, it allows the calculation of consequence cost due to bone fracture. In our study, after estimating the excess medical cost for bone fracture ($37,179), we further factored it into two components: medical treatment costs for bone fracture ($26,033), which captured the costs from the claims associated with the ICD-9 codes for bone fractures, and other excess treatment costs for bone fracture ($11,146), which are the residual of excess medical costs after subtracting the medical treatment costs. Therefore, we believe that we captured the hidden consequence cost of bone fracture. Second, this study allowed estimation of the excess rate of resource utilization due to bone fracture. The approach of estimating the excess long-term care admission rates due to bone fractures seems appropriate. The third unique feature is that the study separated the two groups of patients with early-stage breast carcinoma and bone fractures into those ever hospitalized for bone fracture and those never hospitalized for bone fracture within the 2-year period. This distinction is important, because these two groups incurred substantially different direct costs for bone fracture. The ever-hospitalized group averaged $26,033 in treatment costs for bone fractures, whereas the never-hospitalized group averaged only $1385. The ever-hospitalized group probably had severe bone fractures or new (incident) fractures. The never-hospitalized group may have had less severe bone fractures, which usually do not require hospitalization, or were hospitalized before the 2-year observation period and, thus, needed only some follow-up care. Nevertheless, both groups appear to impose a significant economic burden to society in the U.S. Finally, the distinction between the two groups demonstrated that the biggest determinant of cost is hospitalization for severe bone fractures. Thus, a pharmacologic treatment that could prevent or delay those events probably would be cost effective.

Study Limitations

This study had several limitations. Our prevalence approach study resulted in the estimates of costs of bone fracture incurred by Medicare within a 2-year period. We identified patients with early-stage breast carcinoma and bone fractures by reviewing an individual's record of medical claims; thus, a patient's fracture may have occurred near the beginning of the 2-year period; whereas her first, early-stage breast carcinoma diagnosis may have occurred near the end. In this instance, a nearly 2-year gap would make it problematic to assume that the fracture was associated with early-stage breast carcinoma or its treatment. However, we believe that the impact of such an event is limited. It also is true that the cost of bone fracture should be underestimated compared with an incidence-based study that identified patients from their first diagnosis and had a 2-year follow-up. Most patients with cancer received a diagnosis when the symptoms became apparent, and the disease had progressed (for example, from Stage II); however, if the time between the onset of breast carcinoma and the first diagnosis of early-stage breast carcinoma was long enough, then a bone fracture detected before the first diagnosis of early-stage breast carcinoma still occurred after the onset of the disease itself. Nevertheless, the 2-year window of our prevalence study does not guarantee that the first diagnosis of early-stage breast carcinoma we found was an incident case. For those patients, costs may have been underestimated.

Another possible limitation stems from our exclusion of all patients who died during the study period (1997–1998). Ideally, we should have excluded only patients who died of cancer; however, we did not have information regarding the cause of death. Three additional considerations are related to this exclusion. First, we wanted to examine long-term survivors (that is, those who usually receive endocrine therapies, such as antiestrogens and aromatase inactivators and inhibitors). Second, we wanted 2 full years of observation to make cost and prevalence calculations more straightforward. Third, death for a patient with cancer is an obvious sign of late-stage disease; therefore, to select patients with early-stage breast carcinoma, we deliberately excluded all patients who died. However, we also tested the impact of such omission by adding those deaths and reestimating the treatment costs of bone fracture. After we added 21 qualified patients who died after their fracture and adjusted for the truncated time periods, the treatment cost for bone fracture in patients with early-stage breast carcinoma who were ever hospitalized for bone fracture rose from $26,035 to $28,217. For those who were never hospitalized, the cost increased from $1386 to $1544. We then decided that it was more desirable to keep the conservative estimate, which did not contain those deaths.

A final limitation concerns our use of the SAF Medicare 5% data base. Although this data base is a reliable and convenient source for the estimation of costs, the limitations of its medical claims data for outcomes research have been documented well.11–18 In addition, the definition of bone fracture was based solely on the ICD-9 diagnosis codes in the medical records. This definition would be reasonably precise for women who were ever hospitalized for bone fracture because of the reliability of hospital records and the severity of the symptoms. It may not be as precise for women who were never hospitalized, because some of the fractures were suspected but were never confirmed. Moreover, because Medicare did not cover long-term care, we used the place of service to identify women who ever received services at long-term care institutions as an indication that they had a long-term care admission. Using the place of service to signal a long-term care admission may underestimate the admission rate, because such patients may seek care in other medical settings. Nevertheless, the impact of this issue probably is limited, because the rates of admission to long-term facilities were high, as expected, and our estimates are consistent with those reported in the literature.4

Conclusions

This study provides new evidence that bone fracture in patients with early-stage breast carcinoma is an important public health issue that contributes substantially to health expenditures in the U.S. The current results suggest that bone fracture incurs substantial health expenditures. Therefore, a pharmacologic intervention that could prevent fractures should be cost effective. Our findings indicate that inpatient hospital costs for bone fractures account for only 27% of the total direct costs of bone fracture. Thus, focusing only on inpatient treatment costs for bone fracture may overlook other opportunities for potential cost containment. Nevertheless, our findings of a drastic cost difference between the cohorts of women who were ever hospitalized for bone fracture and women who were never hospitalized for bone fracture suggest that the hospitalization for severe fracture triggers substantial expenditures. Timely treatment of less severe fractures could prevent later high-cost hospitalizations and warrants more attention from the medical community. The lack of a significant difference in medical treatment costs for bone fracture per person between the early-stage breast carcinoma cohort and the cohort of normal women may be explained largely by our focus on women with early-stage breast carcinoma. Two important issues require additional research: 1) the incidence of bone fracture in women who have early-stage breast carcinoma compared with the incidence in an age-matched sample of women without a diagnosis of breast carcinoma and 2) the possible effects of currently available endocrine treatments (such as tamoxifen and aromatase inhibitors and aromatase inactivators) on the bone metabolism of this large patient population. Incidence-based studies that investigate the costs associated with bone fracture in women with different stages of breast carcinoma, as well as head-to-head comparisons among the main endocrine therapies currently available, should provide a wider and more thorough perspective for this important medical and economic issue.

Ancillary