Direct Healthcare Costs of Osteoporosis-Related Fractures in Managed Care Patients Receiving Pharmacological Osteoporosis Therapy

Transcription

1 ORIGINAL RESEARCH ARTICLE Appl Health Econ Health Policy 2012; 10 (3): /12/ /$49.95/0 Adis ª 2012 Springer International Publishing AG. All rights reserved. Direct Healthcare Costs of Osteoporosis-Related Fractures in Managed Care Patients Receiving Pharmacological Osteoporosis Therapy Hema N. Viswanathan, 1 Jeffrey R. Curtis, 2 Jingbo Yu, 3 Jeffrey White, 4 Bradley S. Stolshek, 1 Claire Merinar, 1 Akhila Balasubramanian, 1 Joel D. Kallich, 1 John L. Adams 5 and Sally W. Wade 6 1 Amgen Inc., Thousand Oaks, CA, USA 2 University of Alabama at Birmingham, Birmingham, AL, USA 3 HealthCore, Inc., Wilmington, DE, USA 4 WellPoint Inc., Thousand Oaks, CA, USA 5 RAND Corp., Santa Monica, CA, USA 6 Wade Outcomes Research and Consulting, Salt Lake City, UT, USA Abstract Background: Osteoporosis is a common condition and the economic burden of osteoporosis-related fractures is significant. While studies have reported the incremental or attributable costs of osteoporosis-related fracture, data on the economic impact of osteoporosis-related fractures in commercial health plan populations are limited. Objective: To estimate the direct costs of osteoporosis-related fractures among pharmacologically treated patients in a large, commercially insured population between 2005 and Methods: In this retrospective cohort study, patients were identified from a large, commercially insured population with integrated pharmacy and medical claims. Inclusion criteria were age years; one or more osteoporosis medication claim(s) with first (index) claim between 1 January 2005 and 30 April 2008; and continuous insurance coverage for 12 months pre-index and 6 months post-index. Patients with pre-index Paget s disease or malignant neoplasm; skilled nursing facility stay; combination therapy at index; or fracture 6 months post-index were excluded. A generalized linear model compared differences in 6-month pre-/post-event costs for patients with and without fracture. Propensity score weighting was used to ensure comparability of fracture and non-fracture patients. Generalized estimating equations accounted for repeated measures. Results: The study included patients (2613 with fracture) with a mean (SD) age of 56.4 (4.7) years; 95.9% were female. Mean differences between pre- and post-event direct costs were $US (95% CI 7670, ) for

2 164 Viswanathan et al. patients with vertebral fractures, $US (95% CI , ) for patients with hip fractures, and $US7582 (95% CI 6532, 8632) for patients with other fractures. After adjusting for covariates, osteoporosis-related fractures were associated with an additional $US9996 (95% CI 8838, ; p < ) in direct costs per patient across all fracture types during the 6 months following fracture. Conclusion: Patients with osteoporosis-related fractures were found to incur nearly $US in estimated additional direct healthcare costs in the 6 months post-fracture, compared with patients with no fracture. Reduced fracture risk may lower associated direct healthcare costs. Key points for decision makers Patients with osteoporosis-related fractures incurred nearly $US in estimated additional direct healthcare costs in the 6 months post-fracture, compared with patients with no fracture Increased efforts to reduce fracture risk in large health plans may lower associated direct healthcare costs Introduction Osteoporosis is a common condition affecting an estimated 44 million Americans. [1] A systemic disease characterized by low bone mass, osteoporosis can result in decreased bone strength and increased risk of fractures. [1] In 2000, an estimated 9 million new osteoporotic fractures were reported, of which 1.4 million were clinical vertebral fractures, 1.6 million were hip fractures and 1.7 million were forearm fractures. [2] Osteoporotic fractures can result in serious disability and increased economic costs. [2] The number of medications available for osteoporosis prevention and treatment has expanded over the last decade. Despite the availability of a wide range of therapies with different mechanisms of action, modes of administration, and dosing schedules, under-treatment is common, osteoporosis-related fractures are frequent, and fracture-related costs remain high. [1,3-6] In previous studies, the economic burden of osteoporosis-related fractures has been reported as significant; amounting to estimated projected costs of nearly $US19 billion in the US in [3,7,8] The generalizability of previous studies examining the costs of osteoporosis-related fractures often has been limited by small sample sizes and lack of geographic diversity. [9,10] Previous studies have also been limited by the use of older data or assumptions to create estimation models. [11,12] In addition, most of the existing literature has focused on estimating hip fracture costs, and fewer studies have assessed the costs of vertebral and other fractures. The primary objective of this study was to estimate the incremental direct costs (health plan reimbursements) associated with osteoporosisrelated fractures in a large, geographically diverse, commercially insured, managed care population between 2005 and 2008 using rigorous methodology and a validated fracture identification algorithm. Unlike other recent administrative claims studies that have focused on all patients with specific types of fracture, or fractures in patients with diagnoses of osteoporosis, [4,13,14] this study quantified fracture costs among patients receiving pharmacological osteoporosis therapy. Thus, the costs reported represent the most potentially avoidable costs among patients for

3 Costs of Osteoporosis-Related Fractures 165 whom physicians deemed osteoporosis therapy appropriate. Patients and Methods Data Source This retrospective cohort study used administrative claims data from the HealthCore Integrated Research Database (HIRDÔ) to estimate fracturerelated costs from a health plan perspective. The HIRDÔ contains longitudinal claims data from 14 health plans across the US. It includes linked medical, pharmacy and eligibility data from approximately 33 million covered individuals who received medical care through health maintenance organizations, point-of-service, preferred provider organizations (PPOs) and indemnity plans. The HIRDÔ is updated with fully adjudicated paid claims data on a monthly basis, with most adjudicated medical claims available within 3 months of service rendered date. Most pharmacy data are incorporated into the HIRDÔ within 1 to 2 months after the prescription fill date. Demographic and cost data were complete for all patients in the study population. All data were handled in compliance with the Health Insurance Portability and Accountability Act of 1996 (HIPAA). Patient Selection Patients were identified using medical and pharmacy claims for services incurred from 1 January 2005 to 30 April 2008, defined as the intake period, and follow up extended to 31 October Study patients were required to have one or more pharmacy claims for one or more of the following osteoporosis medications during the intake period: risedronate, alendronate, ibandronate, calcitonin, raloxifene and teriparatide. Zoledronic acid users were excluded due to small numbers and insufficient follow-up data. For each patient, the index date was set to the date of the first claim for a qualifying osteoporosis medication during the intake period. Patients were new to osteoporosis therapy, with no claims for osteoporosis medication in the 12 months prior to the index date; aged 45 to 64 years at index; and required to have 12 months of continuous medical and pharmacy insurance coverage before the index date and 6 months after the index date. Patients with fewer than 12 months of follow up, including those who died during that interval, were excluded. Patients were excluded if they had other metabolic bone diseases or conditions other than osteoporosis that might increase fracture risk (i.e. defined as Paget s disease [International Classification of Diseases, 9th Edition (ICD-9) 731.0] or malignant neoplasm [ICD ] pre-index; or pharmacy claim(s) consistent with bisphosphonate therapy at doses used for Paget s disease during the study period); if they were on combination osteoporosis therapy at index (claims for two or more different osteoporosis agents at index); or if they had claims for skilled nursing facility care at index. Patients who had fracture-related claims within 6 months postindex were also excluded in order to (i) ensure sufficient time for osteoporosis treatments to have a clinical impact, and (ii) limit misclassifying follow-up for prevalent fractures as incident fractures. Cohort Assignment and Fracture Definitions Observation time began 6 months following each person s index date, and patients were assigned to a fracture or non-fracture cohort based on whether they sustained an incident fracture while under observation. Incident fractures were identified with validated algorithms, and classified by skeletal location. [15,16] Fractures could have been treated in either outpatient or inpatient settings. Clinical thoracic and lumbar vertebral fractures required ICD-9 diagnosis codes of , 805.2x, 805.4x, and a radiologic procedure within 30 days prior to the fracture diagnosis. Other vertebral fractures required a procedure code within 30 days prior to the diagnosis and a gap of 120 days from the last similar fracture claim. Fractures of the hip (ICD x and ), pelvis, femur, lower forearm, radius/ulna and humerus required an ICD-9 diagnosis as well as a procedure code for repair within 30 days of the fracture diagnosis. All other fractures (excluding skull, finger or toe) required a fracture diagnosis on an emergency room or inpatient claim, or a gap of 120 days from the last similar fracture

4 166 Viswanathan et al. claim. Incident fractures were categorized into three groups: vertebral fractures (i.e. thoracic and lumbar vertebral fractures); hip fractures; and other fractures (i.e. vertebral fractures other than dorsal/lumbar and fractures at locations other than vertebral and hip). For patients in the fracture cohort, the event date was assigned as the date of the first qualifying post-index fracture. For patients in the nonfracture cohort, the event date from a randomly selected fracture cohort patient was assigned, stratified by a comparable amount of time in the health plan from index to end of follow-up period (see figure I, Supplemental Digital Content [SDC], Fracture and non-fracture patients were first classified by the total time between their index date and the end of their health plan eligibility. The maximum observed eligibility period was split into 40 segments of approximately 30 days each. Each patient was assigned to the segment corresponding to the length of their health plan eligibility. For each fracture patient, the number of days between index date and fracture date was determined and this time interval was used to set the event date for a randomly selected non-fracture patient in the same eligibility segment. This approach ensured that fracture and non-fracture patients had equivalent follow-up periods. All fracture and non-fracture patients with 6 months of followup data after their respective event date were retained for analysis. Patients with fewer than 6 months of continuous health plan enrolment post-event date for any reason, including death, were excluded. Outcome Variables Since the analyses were performed from the perspective of the insurer, health plan allowed amounts were used to quantify the total direct healthcare costs. Healthcare utilization (count) and costs were categorized by service location, including inpatient stay, emergency room visits, office visits, other outpatient care (e.g. home care, outpatient hospital visits), skilled nursing facility, and outpatient pharmacy. All healthcare utilization and cost variables were measured during the 6 months prior to and 6 months after the event date. Costs were inflated to 2009 $US using the Consumer Price Index for medical care. [17] Statistical Analysis Baseline patient characteristics, healthcare utilization and direct costs were summarized and compared for patients in the fracture and nonfracture cohorts. For patients in the fracture cohort, descriptive analyses were performed to assess healthcare utilization and costs by fracture location. Mean and standard deviation (SD) were reported for continuous variables (mean and 95% confidence intervals [CI] for utilization and costs), and frequency and percentage for categorical variables. Kruskal-Wallis tests compared continuous variables and Chi-square tests compared categorical variables. The direct costs of fractures were estimated with a gamma generalized log-linear model (GLM) using a difference-in-differences approach. Specifically, costs incurred in the 6 months postevent were compared with costs incurred in the 6 months prior to the event date for each patient. The resulting differences were compared for the fracture and non-fracture cohorts. The multivariate analysis was inverse probability propensity score-weighted to balance and increase comparability of fracture and non-fracture patients. The standardized incidence ratio method was used to calculate the propensity score weights. This approach reweights the non-fracture cohort to more closely match the fracture cohort. The variables used to generate the propensity for fracture in the fracture cohort were age, sex, region, health plan type, evidence of pre-index bone mineral density assessment, pre-index osteoporosis diagnosis, osteoporosis medication use between index and event date, pre-index fracture, Deyo Charlson Comorbidity Index (DCI) score and several specific comorbidities (such as arthritis [osteoarthritis and rheumatoid arthritis], musculoskeletal pain, hypothyroidism, respiratory disorders, menopausal symptoms, cardiovascular disease, diabetes mellitus, dementia/depression/anxiety/sleep disorders), number of pre-index medications (excluding osteoporosis agents), pre-index total

5 Costs of Osteoporosis-Related Fractures 167 healthcare costs, days from index to event date and length of post-index follow-up. Comorbidities were selected based on a hypothesized clinical relationship to fracture and expected significant impact on direct healthcare costs. Generalized estimation equation methods [18,19] were used to adjust standard errors for the repeated measures correlation of individuals from pre- to post-event periods. Model coefficients were back-transformed to present results in a dollar scale. All analyses were also conducted using females only to estimate the incremental costs of osteoporotic fractures in women. Statistical analyses were performed using Stata version 10.1 SE (StataCorp, College Station, TX). Representatives of the sponsor (Amgen Inc.), HealthCore and WellPoint designed the study. The statistical analysis of the data was conducted according to a pre-specified plan. Results A total of patients met the initial inclusion criteria. Approximately 55% of these patients were excluded for having an event or fracture date within 6 months of the index date, or <6 months of post-event health plan eligibility (Step 8, table I), resulting in a sample of patients for the cost analysis (table I). Selected baseline characteristics are presented in table II. Of these patients, 2613 (5.3%) had evidence of incident fracture(s) 6 months post-index (fracture cohort). The majority of study patients (n = ; 94.7%) had no evidence of fracture during followup (non-fracture cohort). Most patients (95.9%) were female, mean (SD) age was 56.4 (4.7) years, 62.3% were covered through a PPO health plan, and 5.7% had 1 pre-index fracture. In the fracture cohort, the first incident fracture was vertebral in 245 (9.4%) patients, hip in 158 (6.0%) patients and at other locations in 2210 (84.6%) patients. The fracture and non-fracture cohorts differed significantly in pre-index fracture experience (17.5% of patients vs 5.1% of patients, p < ), mean DCI score (0.5 vs 0.3, p < ), number of concurrent medications (8.6 vs 6.5, p < ), and certain comorbid conditions. After applying propensity score weights, patient characteristics were comparable between fracture and non-fracture patients, with the exception of pre-index fracture experience (see table I in the SDC). A statistically higher percentage of patients in the fracture cohort had experienced a preindex fracture compared with patients in the nonfracture cohort (17.5% vs 5.1%, p < ). Direct Costs of Fractures In the fracture cohort, the 6-month unadjusted mean total healthcare costs per patient were $US6446 (95% CI 5831, 7060) before fracture and $US (95% CI , ) after fracture, with an average increase of $US8737 per patient (95% CI 7623, 9851) in the post-fracture period Table I. Patient counts at key points in selection process Step All patients Fracture patients Non-fracture patients Step 1: 1 claim for osteoporosis medication during intake period Step 2: Age requirement 45 years Step 3: 12 months pre-index and 6 months post-index continuous eligibility requirement Step 4a: Exclude patients with Paget s disease or neoplasm during study period Step 4b: Exclude patients with bisphosphonate use for Paget s disease during study period Step 4c: Exclude patients who used a skilled nursing facility at index Step 5: Exclude patients who had multiple index osteoporosis drugs (by names) on index date Step 6: Exclude patients who were on combination osteoporosis medication after index date Step 7: Exclude patients who were 65 years or older at index Step 8: Exclude patients whose event date a was in the first 6 months post-index, and patients whose event date to end-of-eligibility time period was <6 months a Event date is date for first incident fracture or assigned event date.

6 168 Viswanathan et al. (table III). In the non-fracture cohort, the 6-month total healthcare costs per patient were $US3354 (95% CI 3272, 3436) before the assigned event date and $US3372 (95% CI 3285, 3460) after the event date, with no significant change (mean $US18; 95% CI -60, 96) between these time periods. For the fracture cohort, costs in the postfracture period varied by fracture location (table III; figure 1). The unadjusted mean direct costs per patient during the 6 months following the fracture were $US (95% CI , ) for vertebral fractures, $US (95% CI , ) for hip fractures, and $US (95% CI , ) for patients with other fractures. Vertebral fractures and hip fractures were associated with comparatively larger direct costs than other fractures in both the pre- and post-fracture periods (table III). These costs reflect fracture patients Table II. Demographic, clinical and selected treatment characteristics All patients Fracture patients Non-fracture patients Fracture vs non-fracture p-value Patients, n (%) (100.0) 2613 (5.3) (94.7) Number of days for post-index eligibility, mean (SD) 1071 (296) 1135 (266) 1067 (297) < Age [y], mean (SD) 56.4 (4.7) 56.7 (4.6) 56.4 (4.7) < Age group, n (%) y (33.7) 805 (30.8) (33.9) y (66.3) 1808 (69.2) (66.2) Female, n (%) (95.9) 2461 (94.2) (96.0) < Geographic region, n (%) East 9235 (18.6) 551 (21.1) 8684 (18.5) < South (25.4) 696 (26.6) (25.3) Central (20.7) 503 (19.3) 9772 (20.8) West (35.4) 863 (33.0) (35.5) Health plan type, n (%) HMO (25.3) 513 (19.6) (25.6) < POS 3741 (7.5) 187 (7.2) 3554 (7.6) PPO (62.3) 1740 (66.6) (62.1) FFS 463 (0.9) 29 (1.1) 434 (0.9) Other 1950 (3.9) 144 (5.5) 1806 (3.8) Pre-index fracture, n (%) 2875 (5.7) 457 (17.5) 2418 (5.1) < Concomitant non-osteoporosis medications, mean (SD) 6.6 (4.7) 8.6 (6.3) 6.5 (4.6) < DCI score, mean (SD) 0.3 (0.7) 0.5 (1.0) 0.3 (0.7) < Pre-index diagnosis a,n(%) Osteoporosis (53.2) 1751 (67.0) (52.4) < Osteopenia (27.8) 684 (26.2) (27.9) Menopause and menopausal symptoms (25.0) 600 (23.0) (25.2) Arthritis (osteoarthritis and rheumatoid arthritis) (24.7) 904 (34.6) (24.1) < Musculoskeletal pain (22.6) 829 (31.7) (22.0) < Respiratory diseases 9674 (19.5) 672 (25.7) 9002 (19.1) < Hypothyroidism 8961 (18.0) 505 (19.3) 8456 (18.0) a Pre-index diagnoses were identified from any claim with the specified code during the 12 months pre-index. Rates of pre-index diagnoses of dementia, depression, anxiety, sleep disorders; diabetes; upper gastrointestinal disorder; cardiovascular disease; cerebrovascular disease; indication of decreased mobility; kidney disease; nutritional disorders; obesity; Parkinson s disease; alcohol use disorder; bulimia or anorexia were also significantly different between the fracture and non-fracture groups (p < 0.05). DCI = Deyo Charlson Comorbidity Index; FFS = fee-for-service; HMO = health maintenance organization; POS = point-of-service; PPO = preferred provider organization; SD = standard deviation.

7 Costs of Osteoporosis-Related Fractures 169 Table III. Mean unadjusted total direct healthcare costs ($US) per patient 6-month pre-event costs, mean (95% CI) 6-month post-event costs, mean (95% CI) Post-event minus pre-event costs (difference), mean (95% CI) Clinical vertebral fractures n = (8617, ) (21 339, ) (7670, ) Hip fractures n = (5912, ) (19 893, ) (11 690, ) Other fractures n = (5085, 6112) (12 043, ) 7582 (6532, 8632) All fractures combined n = (5831, 7060) (14 031, ) 8737 (7623, 9851) No fracture n = (3272, 3436) 3372 (3285, 3460) 18 (-60, 96) increased utilization of healthcare services (hospitalizations, emergency room visits, office and other outpatient visits, skilled nursing and outpatient prescriptions) during the 6 months after fracture (table IV). Further analysis of the postfracture data indicated that vertebral fracture patients in the highest quartile of post-fracture costs had much greater use of healthcare services, including inpatient and skilled nursing facility care (see table II in the SDC). Using the propensity score-weighted GLM model, the mean costs during the 6 months after the fracture were $US9996 (95% CI 8838, ) per patient higher than the mean per patient costs for those patients with no incident fractures, after adjusting for the clinical differences in the patient populations (see table III in the SDC). Factors associated with higher mean direct costs in the fracture cohort were female gender, living in a south or central region, PPO health plan, higher DCI score, no fracture pre-index, pre-index osteoporosis medication use, pre-index arthritis diagnosis, pre-index Parkinson s disease diagnosis, polypharmacy (more than five concurrent medications used during pre-index) [data not shown]. A simplified linear version of the model had an R- squared statistic of 23%. In subgroup analysis conducted for female patients, the unadjusted mean direct costs per patient during the 6 months following fracture were $US (95% CI , ) for vertebral fractures, $US (95% CI , ) for Median total costs Mean total costs Healthcare costs ($US) Vertebral fractures n = 245 Hip fractures n = 158 Other fractures n = 2210 No fracture n = Fig. 1. Mean and median 6-month unadjusted healthcare costs for fracture patients by fracture location. Healthcare costs in the post-fracture period varied by fracture location, with unadjusted mean direct costs per patient during the 6 months following the fracture of $US for vertebral fractures, $US for hip fractures, and $US for patients with other fractures.

8 170 Viswanathan et al. Table IV. Six-month healthcare costs ($US) for fracture patients and non-fracture patients 6-month post-fracture event total costs (allowable amount) Type of fracture Clinical vertebral fractures n = 245 Hip fractures n = 158 Other fractures n = 2210 No fracture n = Hospitalizations Emergency room visits Office visits Other outpatient visits a Skilled nursing facility Pharmacy Total a Includes laboratory and advanced imaging services. hip fractures and $US (95% CI , ) for women with other fractures. Among female patients, the adjusted mean total costs per patient in the 6-month post-event period were $US9512 (95%CI 8364, ) higher for fracture patients compared with non-fracture patients. Discussion This study, conducted and interpreted from a health plan perspective, demonstrated that patients with an osteoporosis-related fracture incurred, on average, $US in additional direct healthcare costs (i.e. health plan reimbursements) in the 6 months post-fracture compared with patients without a fracture. These incremental costs were incurred through fracture patients increased utilization of all types of healthcare services during the 6 months after fracture, and represent the most potentially avoidable costs for individuals whose physicians decided that pharmacological therapy was appropriate for treating their osteoporosis. Fracture patients not only had higher service utilization and costs in the post-fracture period compared with the 6 months pre-fracture, but their utilization and costs were substantially higher than patients in the non-fracture group before and after the fracture or event date. As expected, utilization was stable in non-fracture patients in the pre- and post-event periods. In our study, 15% of observed incident fractures occurred at the hip and spine. Wide variations have been reported in the literature in the rate of hip and spine fracture as a proportion of incident osteoporosis-related fractures. Two US claims-based studies reported rates of 21% to 34%, [4,14] while a study based on German sickness fund data reported a rate of 42%. [20] The variation in reported rates could be the result of differences in incidence across populations or variations in the methods used to identify fractures. Our algorithm for fracture identification, which is based on a published classification approach, [21] may have reduced the potential for misclassification (i.e. erroneous assignment of prevalent fracture events to incident fracture events) that existed in prior studies. Although the literature indicates that mean costs for hip fractures are typically higher than mean vertebral fracture costs, [4] in our sample the unadjusted mean costs for patients with vertebral fractures were higher than for hip fracture patients in both the pre- and post-fracture periods. By contrast, the median post-fracture costs for vertebral fracture patients were substantially lower than for hip fracture patients ($US vs $US19 223), suggesting that outliers may have skewed the mean costs for vertebral fracture patients upward. Further analysis of the postfracture data revealed that a small number of vertebral fractures resulted in unusually long hospital and skilled nursing facility stays, and therefore these relatively few cases had a disproportionate effect on the total post-fracture costs for this group of patients. In addition, the mean pre-fracture/event costs were more than 50% higher than for hip fracture patients, suggesting that the patients who experienced clinical vertebral fractures during the study period may have been sicker.

9 Costs of Osteoporosis-Related Fractures 171 The comparatively high costs for the vertebral fracture patients may partially result from the ascertainment bias inherent in the identification of vertebral fractures. In an attempt to reduce potential misclassification, fracture identification in this analysis required evidence of an imaging study in addition to a relevant diagnosis code. The relationship between radiographically apparent deformities or morphometric fractures and clinical fractures is not yet fully understood, but a recent study suggests that one-quarter of radiographic vertebral deformities were diagnosed as clinically incident fractures. [22] In general, clinical vertebral fractures are more likely than morphometric fractures to be brought to the physician s attention during routine care, and are also more costly to treat given their greater severity. Since this study is based on claims data, the only vertebral fractures we were able to identify were those for which the patient sought and received medical treatment, and it is unlikely that we captured purely morphometric deformities. The costs of osteoporosis-related fractures have been estimated for hypothetical patient populations using economic models, and retrospective and prospective designs. Previously published estimates of the costs of osteoporosis-related fractures vary due to differences in the overall study design and the size and characteristics of the study population, the types of fractures included, the time periods over which utilization and costs were assessed and different methodologies for the cost comparisons. An early study of the incremental costs of osteoporosis-related fractures matched cases (patients with hip or vertebral fracture) with controls (patients with no fracture) using patient-reported data such as self-perceived health, ability to perform activities of daily life, and living situation, along with patient demographics and physician type. Analysis of 44 casecontrol pairs for incident hip fractures found incremental costs of hip fractures in the year following the event of $US ($US estimated from 1993 Dutch guidelines for cost calculations); analysis of the 42 case-control pairs for incident vertebral fractures found average yearly incremental costs of vertebral fractures (calculated for subjects who had a second radiograph after the case-finding radiograph) of $US1057. [23] This estimate seems consistent with our estimate, allowing for the fact that the earlier study examined costs over a 12-month period and our results are reported in 2009 $US rather than 1993 $US. A larger case-control study used data from a cohort of men and women aged 50 years from Olmsted County, Minnesota, to quantify the direct costs for hip, femur and rib fractures over the first year post-fracture. [9] Using data for 1263 case (fracture) and control (non-fracture) pairs, median total direct medical costs over the initial 12-month evaluation period (pre-fracture) were reported as $US761 for the fracture group and $US625 for the non-fracture group in 1995 $US. [9] In the year post-fracture, costs increased about 5-fold to $US3884 for the fracture group, and remained stable in the non-fracture group at $US712. [9] The incremental median costs were $US for distal femur fractures, and $US for hip fractures. [9] These results seem consistent with the Dutch study and would likely be within the same range as our results if adjustments were made for inflation and the shorter evaluation period used in our study. More recently, researchers have conducted retrospective analyses of large administrative claims databases in an effort to quantify the costs of osteoporotic fractures. Our results are consistent with a case-control study that reported adjusted mean fracture-related costs in the first year post-fracture (occurring between 1 July 2001 and 31 December 2004) of $US15 196, $US6701, and $US6106 for patients aged 50 to 65 years with hip, vertebral and other fractures, respectively. [14] Another recent retrospective claims analysis reported incremental costs per patient over the 12 months post-fracture as $US5381 (standardized to 2006 $US) for privately insured patients with other fractures. [13] When hip fractures were included, the incremental costs per patient were $US5961. In contrast, Christensen et al. [4] report higher mean fracture-related costs for a commercially insured population in the first year postfracture than our results and those reported in the literature: $US (hip); $US (vertebral); $US6434 (non-hip, non-vertebral; standardized to 2006 $US). These studies illustrate the

10 172 Viswanathan et al. difficulties inherent in comparing published fracture cost estimates as the approaches to fracture identification, patient populations, study time period, definition of cost variables and follow-up periods over which costs were assessed all vary considerably. In addition to the size of our data source, the robustness of our study was enhanced by use of the difference-in-difference approach, which helps minimize potential confounding, including the influence of factors that are not explicitly quantified in the data. Our results are generally in line with previous cost-of-fracture studies using claims data from commercial health plans. Variations in cost estimates across studies are likely attributable to differences in data source, patient characteristics, the time period of the study and how fracture-related costs were identified and assessed. Limitations This study relied on claims data for members of employer-sponsored health insurance programmes. Although the study population was geographically diverse and patients were enrolled in a variety of health plan types, our results may only be generalized to other commercially insured populations, which typically enrol younger people. Indeed, the exclusion of patients over age 65 years means that these results cannot be generalized to Medicare plans in which both healthcare service utilization and reimbursement levels would differ from this commercial health plan population. The study also excluded patients who lacked continuous enrolment during the 12-month postindex period and throughout the 6-month postfracture/event date, including patients who died prior to the end of follow-up. The impact of including patients who died during follow-up may have biased the cost estimates, but the direction and magnitude of the bias is difficult to predict because these patients would likely have been more costly than average in the months leading up to death, but would have contributed no additional costs after the date of death. Additionally, we evaluated fracture costs for patients who were treated with osteoporosis medications; these individuals may be healthier than untreated patients, and thus may have different, presumably lower, fracture-related costs. Since the goal of the study was to estimate osteoporosis-related fracture costs, we used claims for osteoporosis medications to identify osteoporosis patients. This is a standard approach for administrative claims data studies and provides more complete and representative sample identification than use of diagnosis codes only. In our sample, for example, only 81% of patients had a formal diagnosis of osteoporosis or osteopenia. In order to maximize generalizability, we did not limit the population to patients with primary osteoporosis as both primary and secondary osteoporosis increase fracture risk. Our approach, however, does exclude patients with untreated osteoporosis by definition and limits the generalizability of our results to treated patients with commercial health insurance. We assumed that osteoporosis could have contributed to any incident fracture in the study population during follow-up, as all patients had to have filled one or more prescription(s) for an osteoporosis medication for inclusion in the study. Hence, some prevalent or non-osteoporosisrelated fractures may have been inadvertently included, although use of a validated fracture identification algorithm should have minimized this type of misclassification. Other limitations inherent in retrospective claims data include potential miscoding in the data and lack of information on certain potential confounders such as smoking rates, diet, family history of fractures, or heavy alcohol use. In addition, laboratory or diagnostic test results, such as bone mineral density results, were not available. Conclusion Among patients receiving pharmacological osteoporosis therapy, osteoporosis-related fractures were associated with an estimated increase of nearly $US in direct healthcare costs per patient in the 6-month post-fracture period compared with patients with no fracture. Increased efforts to reduce fracture risk in large health plans may ultimately lower associated directhealthcarecosts.

11 Costs of Osteoporosis-Related Fractures 173 Acknowledgements This study was funded by Amgen Inc., and employees of Amgen Inc. assisted in the preparation of this paper. BSS, CM, AB, JDK and HNV are employees of and have stock ownership in Amgen Inc. SW and JA are consultants to Amgen Inc.; SW has received remuneration for participation in this study. At the time of this study, JY was an employee of HealthCore Inc., which received a grant to conduct research, and remuneration for preparing the manuscript from Amgen Inc. JRC is a consultant to and has received honoraria from Amgen Inc., Merck, Novartis, Eli Lilly and Roche, and received support from the National Institutes of Health (AR053351). He has also received grants from Novartis, Merck, Proctor & Gamble and Eli Lilly. TJW has no conflicts of interest. Mandy Suggitt assisted in editing this manuscript, with funding from Amgen Inc. The authors would like to thank David Macarios of Amgen Inc. and Rolin Wade, formerly of HealthCore Inc., for their contributions to conceptualizing this study. All authors contributed to the design of the study and interpretation of results, in addition to making substantive contributions to the writing and revision of the manuscript. JY and JA were responsible for statistical methods and data analysis. HV can act as guarantor for the overall content. References 1. National Osteoporosis Foundation. Fast facts [online]. Available from URL: htttp:// [Accessed 2010 Dec 1] 2. Johnell O, Kanis JA. An estimate of the worldwide prevalence and disability associated with osteoporotic fractures. Osteoporos Int 2006 Dec; 17 (12): Burge R, Dawson-Hughes B, Solomon DH, et al. Incidence and economic burden of osteoporosis-related fractures in the United States, J Bone Miner Res 2007 Mar; 22 (3): Christensen L, Iqbal S, Macarios D, et al. Cost of fractures commonly associated with osteoporosis in a managed-care population. J Med Econ 2010; 13 (2): Curtis JR, McClure LA, Delzell E, et al. Population-based fracture risk assessment and osteoporosis treatment disparities by race and gender. J Gen Intern Med 2009 Aug; 24 (8): Song X, Shi N, Badamgarav E, et al. Cost burden of second fracture in the US Health System. Bone 2011 Apr 1; 48 (4): Becker DJ, Kilgore ML, Morrisey MA. The societal burden of osteoporosis. Curr Rheumatol Rep 2010 Jun; 12 (3): Harvey N, Dennison E, Cooper C. Osteoporosis: impact on health and economics. Nat Rev Rheumatol 2010 Feb; 6 (2): Gabriel SE, Tosteson AN, Leibson CL, et al. Direct medical costs attributable to osteoporotic fractures. Osteoporos Int 2002; 13 (4): Orsini LS, Rousculp MD, Long SR, et al. Health care utilization and expenditures in the United States: a study of osteoporosis-related fractures. Osteoporos Int 2005 Apr; 16 (4): Chrischilles E, Shireman T, Wallace R. Costs and health effects of osteoporotic fractures. Bone 1994 Jul-Aug; 15 (4): Ray NF, Chan JK, Thamer M, et al. Medical expenditures for the treatment of osteoporotic fractures in the United States in 1995: report from the National Osteoporosis Foundation. J Bone Miner Res 1997 Jan; 12 (1): Pike C, Birnbaum HG, Schiller M, et al. Direct and indirect costs of non-vertebral fracture patients with osteoporosis in the US. Pharmacoeconomics 2010; 28 (5): Shi N, Foley K, Lenhart G, et al. Direct healthcare costs of hip, vertebral, and non-hip, non-vertebral fractures. Bone 2009 Dec; 45 (6): Baron JA, Lu-Yao G, Barrett J, et al. Internal validation of Medicare claims data. Epidemiology 1994 Sep; 5 (5): Curtis JR, Mudano AS, Solomon DH, et al. Identification and validation of vertebral compression fractures using administrative claims data. Med Care 2009 Jan; 47 (1): Bureau of Labor Statistics. Consumer price index [online]. Available from URL: [Accessed 2010 Mar 23] 18. Hirano K, Imbens G. Estimation of causal effects using propensity score weighting: an application to data on right heart catheterization. Health Serv Outcomes Res Methodol 2001; 2: Rosenbaum P. Model-based direct adjustment. J Am Stat Assoc 1987; 82: Haussler B, Gothe H, Gol D, et al. Epidemiology, treatment and costs of osteoporosis in Germany: the BoneEVA Study. Osteoporos Int 2007 Jan; 18 (1): Warriner AH, Patkar NM, Curtis JR, et al. Which fractures are most attributable to osteoporosis? J Clin Epidemiol 2011 Jan; 64(1): Fink HA, Milavetz DL, Palermo L, et al. What proportion of incident radiographic vertebral deformities is clinically diagnosed and vice versa? J Bone Miner Res 2005 Jul; 20 (7): De Laet CE, van Hout BA, Burger H, et al. Incremental cost of medical care after hip fracture and first vertebral fracture: the Rotterdam study. Osteoporos Int 1999; 10 (1): Correspondence: Dr Hema N. Viswanathan, One Amgen Center Drive, MS 28-3-A, Thousand Oaks, CA , USA. hemav@amgen.com

12 Copyright of Applied Health Economics & Health Policy is the property of ADIS International Limited and its content may not be copied or ed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or articles for individual use.