Direct and Indirect Costs of Non-Vertebral Fracture Patients with Osteoporosis in the US

Direct and Indirect Costs of Non-VertebralFracture Patients with Osteoporosisin the USCrystal Pike,1 Howard G. Birnbaum,1 Matt Schiller,1 Hari Sharma,1 Russel Burge2,3 andEric T. Edgell2

1 Analysis Group, Inc., Boston, Massachusetts, USA

2 Eli Lilly and Company, Lilly Corporate Center, Indianapolis, Indiana, USA

3 University of Cincinnati, Cincinnati, Ohio, USA

Abstract Background:Osteoporosis is a condition marked by low bone mineral density

and the deterioration of bone tissue. One of the main clinical and economic

consequences of osteoporosis is skeletal fractures.

Objective: To assess the healthcare and work loss costs of US patients with

non-vertebral (NV) osteoporotic fractures.

Methods: Privately insured (aged 18–64 years) andMedicare (aged ‡65 years)patients with osteoporosis (ICD-9-CM code: 733.0x) were identified during

1999–2006 using two claims databases. Patients with an NV fracture (femur,

pelvis, lower leg, upper arm, forearm, rib or hip) were matched randomly on

age, sex, employment status and geographic region to controls with osteo-

porosis and no fractures. Patient characteristics and annual healthcare costs

were assessed over the year following the index fracture for privately insured

(n= 4764) and Medicare (n= 48 742) beneficiaries (Medicare drug costs were

estimated using multivariable models). Indirect (i.e. work loss) costs were

calculated for a subset of privately insured, employed patients with available

disability data (n = 1148). All costs were reported in $US, year 2006 values.

Results: In Medicare, mean incremental healthcare costs per NV fracture

patient were $US13 387 ($US22 466 vs $US9079; p < 0.05). The most ex-

pensive patients had index fractures of the hip, multiple sites and femur

(incremental costs of $US25 519, $US20 137 and $US19 403, respectively).

Patients with NV non-hip (NVNH) fractures had incremental healthcare

costs of $US7868 per patient ($US16 704 vs $US8836; p < 0.05). Aggregate

annual incremental healthcare costs of NVNH patients in the Medicare re-

search sample (n= 35 933) were $US282.7million compared with $US204.1

million for hip fracture patients (n= 7997).Among the privately insured, mean incremental healthcare costs per NV

fracture patient were $US5961 ($US11 636 vs $US5675; p < 0.05). The most

expensive patients had index fractures of the hip, multiple sites and pelvis

ORIGINAL RESEARCH ARTICLEPharmacoeconomics 2010; 28 (5): 395-409

1170-7690/10/0005-0395/$49.95/0

ª 2010 Adis Data Information BV. All rights reserved.

(incremental costs of $US13 801, $US9642 and $US8164, respectively). An-

nual incremental healthcare costs per NVNH patient were $US5381

($US11 090 vs $US5709; p < 0.05). Aggregate annual incremental healthcare

costs of NVNH patients in the privately insured sample (n= 4478) were

$US24.1million compared with $US3.5million for hip fracture patients

(n= 255). Mean incremental work loss costs per NV fracture employee were

$US1956 ($US4349 vs $US2393; p < 0.05). Among patients with available

disability data, work loss accounted for 29.5% of total costs per NV fracture

employee.

Conclusion: The cost burden of NV fracture patients to payers is substantial.

Although hip fracture patients were more costly per patient in bothMedicare

and privately insured samples, NVNH fracture patients still had substantial

incremental costs. Because NVNHpatients accounted for a larger proportion

of the fracture population, they were associated with greater aggregate incre-

mental healthcare costs than hip fracture patients.

Background

Osteoporosis is a condition marked by lowbone mineral density and the deterioration ofbone tissue.[1] According to the National Osteo-porosis Foundation, osteoporosis affects over10million North Americans, and an additional34million are considered at risk.[2] Although itcan occur in both men and women, osteoporosisdisproportionately affects females and the elderly.Approximately 80% of those currently diagnosedwith osteoporosis are women, and an estimated55% of those aged ‡50 years in the US eitheralready have osteoporosis or face elevated riskfor the disease.[2]

One of the main clinical and economic con-sequences of osteoporosis is skeletal fractures.Over 1.5million fractures can be attributed toosteoporosis every year, many of which resultfrom incidents that would not normally cause abroken bone in a healthy person. The NationalInstitute of Health estimates that 50% of womenand 25% of men over the age of 50 years in the USwill have an osteoporotic fracture in their re-maining lifetime.[3] Fractures can be extremelydebilitating. Rehabilitation is lengthy and manynever regain pre-fracture mobility, which canhave a major impact on lifestyle and well-being.[4]

For example, decreased functionality often re-sults in an inability to fulfil social roles, which

may lead to mental disorders such as anxiety ordepression.[5]

The economic burden to payers of patientswith osteoporotic fractures is substantial. Thedirect cost burden of all osteoporotic fractures inthe US was estimated at $US16.9 billion in 2005and is projected to increase to $US25.3 billion by2025.[6] Previous studies indicate that healthcarecosts for patients with non-vertebral (NV) frac-tures (e.g. pelvis, femur, lower leg, forearm, upperarm and rib) are particularly expensive.[7,8] Al-though themean incremental costs of NV fracturepatients over controls varies widely by fracturetype and sample selection criteria, hip fracturepatients typically incur the highest costs duringthe year following a fracture.[9,10] Annual esti-mates of total medical expenditures per hip frac-ture patient range from $US13 240 to $US30 400in 2006 values.[9-13] The limited evidence on theeconomic impact of other NV fracture typessuggests that patients with NV fractures at non-hip sites incur substantial costs as well. Estimatesrange from $US3000 to $US16 800 per patientfor the year following a NV non-hip (NVNH)fracture, with patients having femur, pelvis orlower leg fractures at the high end of this costrange.[10,12]

Most studies of NV fractures have relied on datafrom before 2003; as a result they may no longeraccurately depict the healthcare and economic

396 Pike et al.

ª 2010 Adis Data Information BV. All rights reserved. Pharmacoeconomics 2010; 28 (5)

outcomes of a rapidly expanding at-risk popula-tion. As of 2008, there were over 36millionMedicare beneficiaries aged ‡65 years, and as the‘baby boom’ generation nears retirement, thisfigure is projected to increase to 62million by2025.[14,15] Additionally, there were over 30mil-lion employees aged 55–64 years in 2005, and theUS Census Bureau estimates that this figure willincrease to more than 41 million in 2025.[16] Theconsequences of osteoporotic fractures in theworkplace have not been closely examined; norhave costs been compared between the ‡65 yearsand <65 years populations.

The primary purpose of this research was toassess the health outcomes and economic burdenof patients with NV fractures in the US for bothprivate payers (aged 18–64 years) and Medicare(aged ‡65 years) using administrative claimsdata. In particular, it aimed to update the existingliterature[5-12,17-21] on the direct costs (i.e. medicaland prescription drugs) of NV fracture patientsduring the year following an incident fracture andexpand on it, by describing variations in healthoutcomes and costs among different fracture typesand patient populations (i.e. Medicare vs privatelyinsured). While much has been written about thehealthcare characteristics and costs of patients withhip fractures in the US and Europe,[22-24] relativelylittle is known about patients with other NV frac-tures. The study also estimated indirect work losscosts (medically related absenteeism and dis-ability) for patients with NV fractures – an areafor which there is little existing research.

Methods

Data

This study used de-identified administrativeclaims selected from two databases for servicesincurred during 1999–2006, i.e. an employerclaims database and the Medicare StandardAnalytic Files. The employer database coversover eight million privately insured beneficiaries(including employees, spouses and dependents)from 40 companies that operate nationwide, andincludes a broad array of job classifications andindustries (e.g. energy, financial services, food

and beverage, manufacturing and telecommunica-tions). The data contain information on demo-graphics, monthly enrolment history and medicaland prescription drug claims for all beneficiaries.Short- and long-term disability claims reportingdates of work loss and benefits paid are availablefor employees in a subset of these companies. TheMedicare database is a 5% random sample of allMedicare beneficiaries containing demographicand enrolment information and medical claimssimilar to the privately insured database. ForMedicare, the date of medical service refers to thequarter in which a patient was treated, and nopharmacy claims are available.

Sample Selection

The selection process is summarized in figure 1.A preliminary set of patients with osteoporosis(International Classification of Diseases, NinthRevision, Clinical Modification [ICD-9-CM]code: 733.0x) was identified from each database.Patients were restricted to ages 18–64 years forthe privately insured population and to ages‡65 years for the Medicare population. Patientswith diagnoses at any time in their claims historyfor other conditions that elevate fracture riskwere excluded. Specifically, patients with diagnosesfor metastatic cancer (196.xx–199.xx), bone cancer(170.xx), multiple myeloma (203.xx) and Paget’sdisease of bone (731.0x) were excluded, as werethose who used certain cancer drugs (anastrozole,letrozole, exemestane, leuprorelin, goserelin, trip-torelin, nafarelin, histrelin, deslorelin, flutamide,bicalutamide, nilutamide, cetrorelix, abarelix andganirelix) or underwent surgical castration.[25-27]

To focus on patients with NV fractures, patientswith fractures of the vertebrae (733.13, 805.xx–806.xx), trunk (809.xx), face/skull (800.xx–804.xx),finger (816.xx), toe (826.xx) and unspecified sites(829.xx) were also excluded. Patients with frac-tures of the foot and ankle were included in thesample as there is evidence that these fracturesmay often be osteoporotic.[28,29] These patientswere then classified into two mutually exclusivecohorts: NV fracture patients and potential con-trols with osteoporosis. The NV fracture cohortwas constructed by identifying patients with at

Costs of Osteoporosis Non-Vertebral Fracture Patients 397


least one NV fracture claim (733.11–733.12,733.14–733.16, 807.xx–808.xx, 810.xx–828.xx).The remaining patients were categorized as po-tential controls.

To be included in the final sample, only pa-tients with fractures occurring within 180 days(two quarters for Medicare patients) prior tothe first osteoporosis diagnosis or anytime there-after were included. Fractures occurring up to6 months prior to a patient’s first osteoporosisdiagnosis were considered osteoporotic, becauseosteoporosis often goes undiagnosed until after afracture occurs.[30] The first osteoporotic fracturein the claims history meeting these criteria wasconsidered the index fracture. Patients who had amedical claim with an E-code for trauma within30 days (one quarter forMedicare patients) of theindex fracture were excluded, because these frac-tures may not have been caused by osteo-porosis.[6,7,17] Finally, to reduce the likelihood

that a follow-up visit for a previous fracture wasselected as the index fracture, patients were re-quired to have no other claims for any NV frac-ture in the 6 months (two quarters for Medicarepatients) prior to the index fracture.

These NV fracture patients were then matchedone-to-one to controls with osteoporosis butwithout any fractures. Note that this approachprovides estimates of the incremental cost per NVfracture patient rather than the incremental costof the fracture itself. Privately insured patientswere matched on age (–3 years), sex, geographicregion, employment status and whether they haddisability data available (data on race were un-available). Medicare patients were matched onage, sex, geographic region and race. Controlswere given the same index date as their match,and all patients and controls were required tohave 12 months’ continuous eligibility before andafter the index date.

All patients, 1999–2006(Medicare n = 2 846 792; privately insured n = 8 434 913)

NV fracture patients(Medicare n = 125 147; privately insured n = 13 712)

Osteoporosis patientsPatients (aged 18−64 years for privately insured; aged ≥65 years for Medicare) with an osteoporosis

diagnosis excluding those with metastatic or bone cancer; specified cancer drug treatments;multiple myeloma; Paget's disease of bone; orchiectomy; surgical castration; and fractures to the vertebrae,

trunk, face, skull, fingers, toes or unspecified sites (Medicare n = 454 061; privately insured n = 72 511)

Potential control patients(Medicare n = 328 914; privately insured n = 58 799)

Identified patients with NV fractures within 180 daysbefore first osteoporosis diagnosis or anytime

thereafter, no severe trauma at time of fracture anda clean period prior to the fracture

Study samplePatients and controls with continuous eligibility (randomly matched 1:1 based on age, sex, geographic

region and employment status) [Medicare n = 48 742; privately insured n = 4764]

Subset of NV, non-hip fracture patients and controlsExcludes patients with a hip fracture on their index date (Medicare n = 35 933; privately insured n = 4478)

Fig. 1. Sample selection flowchart. NV = non-vertebral.

398 Pike et al.


Analysis

The analyses and results for the privately in-sured and Medicare samples were conducted andreported separately. Because we treat these sam-ples as completely independent groups, the pre-valence, utilization and costs results provideestimates of the outcomes for the two given pop-ulations, separately. Co-morbidities and medicalresource utilization were assessed descriptivelyover the 12-month period following the indexdate (the four quarters following the index quar-ter for Medicare patients). The Charlson Co-morbidity Index (CCI)[31,32] score was calculatedalong with the proportion of patients and con-trols with CCI co-morbidities and osteoporosis-related co-morbidities that were identified fromthe literature.[7] For resource utilization, visitswere grouped by place of service (e.g. inpatient,outpatient, ED, long-term care) and type ofservice (e.g. radiology, rehabilitation). The pro-portion of patients with at least one visit andthe mean number of visits per patient werecalculated, as were the mean length of stay forinpatient and long-term care visits. For the pri-vately insured population, National Drug Codeswere used to identify the proportion of patientsusing drugs that were osteoporosis-related (e.g.selective estrogen receptor modulators, bisphos-phonates) or fracture-related (e.g. opioids, para-cetamol [acetaminophen]).

There are many alternative ways to conductcost-of-illness studies;[33] as a result, two differentmethods were used in this study (i.e. matched con-trol incremental costs, and regression-adjusted in-cremental costs), with a particular emphasis on thematched control method of assessing incrementalcosts of the fracture patient. Direct costs (medicaland prescription drug costs for the privately insuredand medical costs only for Medicare) were assesseddescriptively over the study period based on actualpayments to healthcare providers from third-partypayers. Costs related specifically to the index frac-ture were also estimated. Index fracture costs weredefined as the costs for claims with a fracture-related ICD-9-CM,Healthcare Common ProcedureCoding System (HCPCS) or Common Proce-dural Terminology (CPT) code as well as the

fracture-related drug costs (privately insuredsample only) incurred during the 6 months (twoquarters for Medicare) following the index date.Finally, regression-adjusted costs were also esti-mated for privately insured patients and matchedcontrols (described further in the Direct Costssection below). All costs were inflated to $US,year 2006 values, based on the Consumer PriceIndex for Medical Care. These results were re-ported for the NV and NVNH groups as well asfor subgroups based on index fracture site.

For the privately insured sample, work loss costs(i.e. employer payments for disability for workabsences that generally exceed 5 days and medi-cally related absenteeism for illnesses that do notmeet disability requirements) were estimated forthe subset of employees with disability data.Disability days and costs were based on actualdisability claims and employer payments forshort- and long-term disability. Days of medi-cally related absenteeism were imputed based onmedical services use (i.e. each hospitalization dayor ED visit counted as a full day of lost work,while an outpatient visit counted as half a day oflost work). Absenteeism costs were then esti-mated by multiplying the number of missed daysby each employee’s daily wage (based on actualemployee wage data). Medically related ab-senteeism did not include days with medical ser-vices occurring during a period of disability, butdid include the qualifying days missed from workpreceding the start of the disability period. Due tosample size constraints, work loss costs were re-ported only for the NV and NVNH groups.

Statistical Analysis

For the descriptive results, McNemar testswere used for the pair-wise comparison of cate-gorical variables, and non-parametric Wilcoxonsigned-rank tests were used for the pair-wisecomparison of continuous variables such as costsand outpatient visits. p-Values < 0.05 were re-ported as statistically significant. All statisticalanalyses were performed using SAS version 9.1(SAS Institute, Cary, NC, USA).

To test the robustness of the matching andensure that the descriptive analyses controlled forbaseline differences between the two groups,



multivariable modelling was used to estimatetotal direct costs for the privately insured popu-lation. Generalized linear regression models witha log-link functional form and Gamma distribu-tion were used to estimate direct costs. In-dependent variables included in the model wereindex fracture type, demographics (e.g. age, age-squared, sex and geographic region), baselineco-morbidities (all osteoporosis-related conditionsand the CCI) and baseline drug treatments (allosteoporosis-related and fracture-related drugs).

Because the Medicare data do not include pre-scription drug claims, multivariable modelling wasalso used to approximate drug costs for these pa-tients (recognizing that Medicare Part D coveragebegan in 2006). Drug costs were estimated for pri-

vately insured fracture patients and controls aged‡65 years. These estimates were then extrapolatedto theMedicare population using a two-partmodelof drug costs prior to 2006 (for whom drug costswere available, as they were covered by their em-ployer-paid insurance benefit). The first part was alogistic model to estimate the probability of a pa-tient having any drug costs, and the second partwas a generalized linear model to estimate the drugcosts conditional on costs being positive. Bothmodels were estimated using fracture type, demo-graphics, study period osteoporosis-related co-morbidities and CCI, and study period medicalresource use as independent variables. Once themodel was specified, the coefficients were appliedto the data (i.e. indicators for independent variables)

Table I. Demographic and co-morbidity characteristics for patients and controls with non-vertebral fractures, 1999–2006

Parameter Privately insured (age 18–64 y) [n =4764] Medicare (age ‡65 y) [n = 48 742]

patients controlsa patients controlsa

Demographics

Age (y) [mean (SD)] 56.5 (6.2) 56.7 (6.3) 78.5 (7.7) 78.3 (7.6)

Female (%) 94.5 94.5 92.2 92.2

Employed (%) 29.3 29.3 NA

Geographic region (%)

midwest 36.6 36.6 45.8 45.8

south 26.1 26.1 13.9 13.9

northeast 25.0 25.0 21.2 21.2

west 12.3 12.3 17.7 17.7

unknown 0.0 0.0 1.4 1.4

Co-morbidities

Osteoporosis-related co-morbidities (%)

arthralgia 41.6 15.9* 41.5 22.8*

hypertension 32.3 28.6* 72.8 68.7*

chest pain 18.5 12.5* 25.8 20.6*

osteoarthritis 18.1 11.1* 46.0 35.1*

malaise and fatigue 16.4 11.1* 30.2 25.6*

abdominal pain 14.0 11.9* 20.1 16.9*

lumbago 12.8 9.3* 14.7 12.6*

urinary tract infection 10.8 8.4* 29.3 20.8*

dyspnoea and respiratory abnormalities 9.9 6.8* 23.7 19.2*

senile cataract 7.2 6.4 26.1 27.3*

fibromyalgia 6.4 4.9* 4.9 4.3*

coronary atherosclerosis 5.1 3.7* 25.5 20.9*

Charlson Co-morbidity Index [mean (SD)] 0.8 (1.3) 0.5 (0.9)* 2.0 (2.1) 1.5 (1.9)*

a Osteoporosis patients with fractures were matched to controls with osteoporosis but no fractures based on age (–3 years), sex,

employment status and geographic region.

NA= not applicable; * p <0.05 patients vs controls for each sample.

400 Pike et al.


for Medicare patients in order to predict theirdrug costs. To validate the model, a randomsample of privately insured patients was excludedfrom the model. Then, the model was used topredict drug costs for this excluded subset, andthe predicted drug costs were compared with ac-tual costs. This analysis assumed that all drugcosts were either paid by Medicare or supple-mental insurance (i.e. it did not attempt to ac-count for the Medicare Part D ‘doughnut hole’).

Results

Patient Characteristics

The mean age of NV fracture patients andcontrols was 56.5 and 56.7 years, respectively, forthe privately insured sample and 78.5 and 78.3years, respectively, for the Medicare sample.Age did not vary substantially by fracture type.The majority of fracture patients were female(privately insured 94.5%; Medicare 92.2%) [table I].

Table II. Resource use for patients (pts) and controls with non-vertebral fractures, 1999–2006

Parameter Privately insured (age 18–64 y) [n =4764] Medicare (age ‡65 y) [n =48 742]

pts controlsa pts controlsa

Medical resource use (% of pts)

Outpatient 99.8 96.2* 99.5 95.8*

radiology 97.0 79.3* 91.5 71.7*

primary care 83.5 77.6* 80.6 79.0*

outpatient hospital 70.0 50.2* 37.9 29.3*

orthopaedic specialist 56.3 12.1* 47.3 10.6*

rehabilitation 50.4 28.2* 36.0 14.2*

other 91.4 84.2* 97.9 92.3*

ED 52.7 19.6* 68.5 27.2*

Hospital inpatient 24.0 9.6* 54.6 23.1*

Long-term care 2.0 0.4* 32.5 10.8*

Other 73.2 60.6* 84.4 70.9*

Prescription drug use (% of pts)

Osteoporosis-related drug use 67.3 61.8* NA NA

alendronate 27.9 21.4* NA NA

estrogen 27.0 28.1 NA NA

prescription calcium 16.1 14.2* NA NA

risedronate sodium 11.5 9.7* NA NA

raloxifene 6.2 6.4 NA NA

ibandronic acid 1.3 1.1 NA NA

teriparatide 1.0 0.3* NA NA

Other osteoporosis-related drug useb 4.6 2.7* NA NA

Fracture-related drug use (% of pts) 68.8 44.7* NA NA

paracetamol (acetaminophen) 51.6 23.2* NA NA

opioids 51.6 22.6* NA NA

NSAIDs (includes cyclo-oxygenase-2 inhibitors) 36.1 25.1* NA NA

systemic corticosteroids 18.5 13.5* NA NA

skeletal muscle relaxants 14.0 9.8* NA NA

tramadol 7.6 3.9* NA NA

a Osteoporosis pts with fractures were matched to controls with osteoporosis but no fractures based on age (–3 years), sex, employment

status and geographic region.

b Other osteoporosis-related medications included calcitonin, etidronate disodium, prescription vitamin D and zoledronic acid.

NA= not applicable; * p <0.05 patients vs controls for each sample.



NV fracture patients had significantly higherrates of co-morbidities than controls during thestudy period. Among the privately insured, theco-morbidities with the greatest differences inprevalence rates were arthralgia (41.6% vs 15.9%;p< 0.05), osteoarthritis (18.1% vs 11.1%; p < 0.05)and chest pain (18.5% vs 12.5%; p< 0.05). Similarly,Medicare patients had the greatest differencesbetween NV fracture patients and controls in theprevalence of arthralgia (41.5% vs 22.8%; p< 0.05),osteoarthritis (46.0% vs 35.1%; p < 0.05) and ur-inary tract infections (29.3% vs 20.8%; p< 0.05).Most osteoporosis-related co-morbidities as wellas those in the CCI were significantly more pre-valent among fracture patients than controls inboth samples (table I). Patient morbidity, asmeasured by the CCI, varied somewhat by frac-ture type. Among fracture patients in the Medi-care sample, the mean CCI was 2.4 for hip andpelvis fracture patients, 1.8 for lower leg fracturepatients and 1.6 for forearm fracture patients.Privately insured hip and femur fracture patientshad CCIs of 1.6 and 1.4, respectively, whereaslower leg and forearm fracture patients had aCCI of 0.9.

Treatment Patterns

Table II reports the medical and drug resourceuse for NV fracture patients and controls in bothsamples. In both privately insured and Medicarepopulations, all measures of medical resource usefor NV fracture patients were significantly higherthan for controls. More than twice as many frac-ture patients had an inpatient visit during theyear following their fracture (privately insured24.0% vs 9.6%; p< 0.05; Medicare 54.6% vs 23.1%;p< 0.05) and almost five times more fracture pa-tients visited an orthopaedic specialist (privatelyinsured 56.3% vs 12.1%; p< 0.05; Medicare 47.3%vs 10.6%; p< 0.05). In addition, more than threetimes as many Medicare fracture patients visiteda long-term care facility during the year followingtheir fracture (32.5% vs 10.8%; p < 0.05). Resultswere generally consistent in the privately insuredsample.

Medical resource use varied greatly betweenpatients with different fracture types. In general,patients with more severe fractures had higherresource utilization. For example, 31.2% ofMedi-care patients with hip fractures began long-term

Table III. Mean (SD) annual direct costs ($US, year 2006 values) for patients and controls with non-vertebral fractures, 1999–2006

Parameter Privately insured (age 18–64 y) [n = 4764] Medicare (age ‡65 y) [n = 48742]

patients controlsa patients controlsa

Total 11 636 (19 918) 5 675 (11 290)* 22466 (25 906) 9 079 (15 147)*

Medical 9 121 (18 508) 3 939 (10 147)* 19488 (25 282) 6 497 (14 489)*

hospital inpatient 3 120 (12 368) 1 012 (7 154)* 9 711 (17 260) 2 932 (10 725)*

long-term care 105 (1605) 12 (459)* 3 819 (8352) 630 (3390)*

ED 324 (980) 100 (530)* 413 (1150) 119 (430)*

outpatient 5 014 (9328) 2 584 (5245)* 4 914 (6000) 2 480 (4107)*

other 558 (2040) 232 (1172)* 630 (2182) 336 (1303)*

Drugb 2 514 (4315) 1 735 (3247)* 2 979 (1443) 2 583 (1223)*

osteoporosis-relatedc 370 (561) 313 (431)* NA NA

fracture-relatedd 276 (1216) 144 (707)* NA NA

other 1 869 (3741) 1 278 (2932)* NA NA

a Osteoporosis patients with fractures were matched to controls with osteoporosis but no fractures based on age (–3 years), sex,

employment status and geographic region.

b Medicare drug costs were imputed based on the privately insured patients over age 65 y, using generalized linear models.

c Osteoporosis-related medications included bisphosphonates, calcitonins, estrogen and SERMs, prescription calcium and prescription

vitamin D.

d Fracture-related medications included paracetamol, corticosteroids, muscle relaxants, NSAIDs and opioids.

NA= not applicable; SERMs= selective estrogen receptor modulators; * p <0.05 patients vs controls for each sample.

402 Pike et al.


care in the quarter in which their fracture oc-curred compared with <10% of patients withfractures of the lower leg, forearm, upper armand ribs. Additionally, more than 70% of Medi-care patients with fractures of the hip, femur,pelvis and multiple sites were hospitalized in aninpatient setting during the study period com-pared with <40% of patients with fractures of theforearm, upper arm, lower leg and ribs.

Among the privately insured, prescriptiondrug use was also higher for fracture patientsthan for controls. Significantly more NV fracturepatients used osteoporosis-related drugs thancontrols (67.3% vs 61.8%; p< 0.05). The onlydrugs for which the proportion of patients fillingprescriptions was not significantly different be-tween patients and controls were estrogen (27.0%vs 28.1%; p = 0.20), ibandronic acid (1.3% vs1.1%; p= 0.34) and raloxifene (6.2% vs 6.4%;p= 0.77). Fracture-related drug use was also sig-nificantly higher for NV fracture patients thanfor controls (68.8% vs 44.7%; p< 0.05). In parti-cular, twice as many fracture patients as controlsused paracetamol (51.6% vs 23.2%; p < 0.05)

and opioids (51.6% vs 22.6%; p < 0.05). The useof NSAIDs was also more common (36.1%vs 25.1%; p < 0.05). Drug use did not vary sub-stantially by fracture type.

Direct Costs

Direct costs were significantly higher for NVfracture patients in both samples (table III).

In the privately insured sample, the mean an-nual incremental direct cost per patient was$US5961 ($US11 636 vs $US5675; p< 0.05). In-patient costs accounted for the most substantialportion of incremental direct costs (35%), whiledrug costs accounted for 13%. The multivariablecost estimates confirmed that privately insuredpatients with an NV fracture were significantlymore expensive than controls with osteoporosisand that pre-existing co-morbid conditions werenot driving the difference in total costs betweenfracture patients and controls. After controllingfor baseline and demographic differences usingthe two-part model, NV fracture patients hadestimated incremental costs of $US5260 ($US11265

Table IV. Direct healthcare costs ($US, year 2006 values) for privately insured patients and controls aged <65 years estimated using three

multivariate models,a 1999–2006

Parameter Average cost Standard deviation Cost difference between

patients and controls

Model 1: NV fracture patients (n =4764)

Patients 11 265 8901 5 260*

Model 2: NVNH fracture patients (n =4478)b

Patients 10 718 8872 4 658*

Model 3: fracture patients by type (n=4764)

Forearm (n= 801) 11 306 9403 5 256*

Upper arm (n= 416) 11 119 9247 5 068*

Ribs (n =608) 8 428 7009 2 377*

Hip (n= 255) 18 613 15 479 12562*

Pelvis (n =80) 12 012 9989 5 961*

Femur (n = 98) 10 563 8785 4 512*

Lower leg (n= 2036) 10 353 8610 4 302*

Multiple fractures (n =470) 15 750 13 099 9 700*

a Generalized linear models were estimated controlling for age, sex, employment status and geographic region as well as baseline co-

morbidities (Charlson Co-morbidity Index and osteoporosis-related), baseline drug use (osteoporosis-related and fracture-related), and

fracture type.

b The NVNH fracture group excludes patients with either a hip fracture as their index fracture or multiple fractures, of which one was a hip

fracture, on their index date.

NV = non-vertebral; NVNH = non-vertebral, non-hip; * p< 0.05.



vs $US6005; p< 0.05) [table IV]. Factors thatsignificantly impacted the privately insured directcosts included having an NV fracture, being male,and having osteoporosis-related co-morbiditiesduring the baseline period (for the coefficientsfrom the two-part model, see table A1 in the Sup-plemental Digital Content 1, http://links.adisonline.com/PCZ/A67).

In the Medicare sample, direct costs (medicalcosts plus imputed drug costs) were higher than inthe privately insured patients for both fracturepatients and controls (table III). The mean an-nual incremental direct cost per NV fracturepatient was $US13 387 ($US22 466 vs $US9079;p< 0.05). Inpatient visits and long-term care werethe cost drivers forMedicare, accounting for 51%and 24% of incremental direct costs, respectively.Prescription drugs were responsible for only 3%of incremental direct costs (see table A2 in theSupplemental Digital Content for a median andinter-quartile range for Medicare direct costs).

Factors that were significantly associated withhigher drug costs included age, having certain co-morbidities during the study period (e.g. coronary

atherosclerosis, hypertension, lumbago), and hav-ing at least one inpatient visit during the studyperiod. Certain index fractures (e.g. hip, femur,multi-ple sites) were significantly associated with lowerstudy period drug costs after controlling for otherfactors such as demographics, co-morbiditiesand physician visits (see table A3 in the Supple-mental Digital Content for point estimates ofdirect costs).

Figures 2 and 3 illustrate the incremental directcosts by fracture type for the privately insuredand Medicare patients. The three most expensiveprivately insured patient groups were those withhip, multiple and pelvis fractures. Incrementaldirect costs for these groups were $US13 801($US18 946 vs $US5145; p< 0.05), $US9642($US14 809 vs $US5167; p < 0.05) and $US8164($US13 296 vs $US5132; p < 0.05), respectively.In the Medicare sample, hip, multiple and femurfracture patients were the most expensive, withincremental direct costs of $US25 519 ($US35 637vs $US10 118; p< 0.05), $US20 137 ($US29 617vs $US9479; p < 0.05) and $US19 403 ($US29 581vs $US10 178; p < 0.05), respectively.

0

2

4

6

8

10

12

14

16

NV(n = 4764)

NVNH(n = 4478)

Hip(n = 255)

Multiple sites(n = 470)

Pelvis(n = 80)

Upper arm(n = 416)

Lower leg(n = 2036)

Forearm(n = 801)

Femur(n = 98)

Ribs(n = 608)

Exc

ess

cost

s ($

US

× 1

000)

DrugsMedical

Fig. 2. Excess direct costs (year 2006 values) for privately insured patients and controls with non-vertebral (NV) fractures, 1999–2006(n= 4764). Note that lower leg fractures include foot, ankle, tibia, fibula and patella; forearm fractures include hand, wrist, radius and ulna;upper arm fractures include humerus, clavicle and scapula. NVNH =non-vertebral, non-hip.

404 Pike et al.


Hip fracture patients (n = 255) in the privatelyinsured sample had aggregate incremental annualdirect costs of $US3.5million to employers com-pared with $US24.1million for NVNH patients(n = 4478). In the Medicare sample, aggregateincremental annual direct costs for hip fracturepatients (n = 7997) were $US204.1million com-pared with $US282.7million for NVNH patients(n = 35 933).

Work Loss Costs

Among privately insured employees, there wasa substantial work loss cost burden associatedwith NV fractures. Nearly three times morefracture patients received disability benefits dur-ing the study period (18.0% vs 6.5%; p < 0.05)compared with controls and their disability peri-ods tended to last more than twice as long onaverage (16.0 vs 6.5 days; p< 0.05). Additionally,fracture patients had more days of medically re-lated absenteeism than controls (13.4 vs 8.6 days;p< 0.05). The additional 2 weeks of missed work

per year among fracture patients resulted inannual work loss costs that were nearly twiceas high as for controls ($US4349 vs $US2393;p < 0.05). For employees with NV fractures, workloss costs accounted for 29.2% of total costs(figure 4). The annual incremental work loss costsof NVNH fracture patients were nearly identical($US4236 vs $US2361; p < 0.05).

Discussion

This study analysed administrative claimsdata to quantify the economic burden per osteo-porotic NV fracture patient paid by private pay-ers and Medicare, with special attention to costdifferences between patients with fractures atdifferent sites as well as the cost impact of NVfractures in the workplace. Both descriptive andregression-adjusted results showed that patientswith NV fractures are significantly more costlyand have significantly worse outcomes thanosteoporosis patients without fractures. From a

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

NV(n = 48 742)

NVNH(n = 35 933)

Hip(n = 7997)

Multiple sites(n = 11 977)

Femur(n = 744)

Pelvis(n = 2022)

Upper arm(n = 4346)

Ribs(n = 6008)

Lower leg(n = 10 137)

Forearm(n = 6291)

Exc

ess

cost

s ($

US

× 1

000)

DrugsMedical

Fig. 3. Excess direct costs (year 2006 values) for Medicare patients and controls with non-vertebral (NV) fractures, 1999–2006 (n= 48 742).Note that lower leg fractures include foot, ankle, tibia, fibula and patella; forearm fractures include hand, wrist, radius and ulna; upper armfractures include humerus, clavicle and scapula. NVNH = non-vertebral, non-hip.



healthcare perspective, NV fracture patientsspend more days in the hospital, have more phy-sician visits and receive more prescription drugs.NV fractures also affected employees, causingthem to miss more days of work and have higherrates of disability than osteoporosis patientswithout fractures, which resulted in increasedcosts to employers.

Not surprisingly, NV fractures imposed agreater burden on the Medicare (‡65 years) pa-tients than on the privately insured (<65 years)population. Medicare patients were more se-verely ill, had higher medical resource use, andconsequently had higher costs. In particular, theyhad substantial incremental long-term care costs,which is especially notable sinceMedicaid is oftenthe primary payer for long-term care. None-theless, the cost burden among the privately in-sured was still substantial. Future research couldexamine the differences in cost burden betweenprivately insured and Medicare patients to ad-dress cost of long-term care, especially, includingwhether age is a significant contributing factor.

These results reaffirm the previous findingsthat hip fractures are the most expensive type offracture on a per-person basis.[9,10] They also

suggest that other fractures such as pelvis andfemur are associated with substantial resourceutilization and costs. While the sample of femurfracture patients in the privately insured popula-tion was small (n = 98), femur fractures were thethird most expensive fracture in the larger Medi-care sample (n = 744), suggesting that the costs offemur fracture patients are an important burden.It is notable that hip fracture patients did notmake up the majority of the total cost burden ofNV fracture patients in either sample. In theprivately insured sample, hip fracture patientsaccounted for only 13% of all aggregate incre-mental direct costs. In the Medicare sample,where hip fractures are more prevalent and morecostly, NVNH fracture patients still accountedfor 58% of the $US487million in total incre-mental direct costs per year of the patients inthis sample. The relatively high prevalence andsubstantial costs of patients with NVNH frac-tures places an economic burden on public andprivate payers that is not widely recognized in theliterature.

The results of this study are generally consis-tent with the existing published literature. Orsiniet al.[10] reported mean annual medical costs of$US19 647 and incremental costs of $US9401 perfracture patient, whereas our study reports$US19 488 in total medical costs and $US12 991in incremental costs for the Medicare population(year 2006 values). Our study reported meanmedical costs of $US9121 and incremental costsof $US5182 for the privately insured sample,which also appear broadly consistent with Orsiniet al.[10] Orsini et al.[10] used a pre/post metho-dology that compares pre-period costs with studyperiod costs, whereas we used a matched controldesign. In addition, the mean age in their samplewas 71 years, whereas the mean age in the Medi-care sample in our study was 79 years. Both factorscould lead to the differences in the cost estimates.

Rousculp et al.[34] published cost estimates foraMedicaid population. They reportedmean annualmedical costs per fracture patient of $US27 081,which were higher than the mean costs for theMedicare population in this study. However,Rousculp et al.[34] found that incremental costsper fracture patient were $US10 358, which are

Disability$US1199 (17.9%)

Medical$US4129 (61.7%)

Drugs$US606(9.1%)

Medically relatedabsenteeism

$US757(11.3%)

Fig. 4. Excess costs (year 2006 values) for privately insuredemployees with non-vertebral fractures by cost type, 1999–2006(n= 1148).

406 Pike et al.


slightly lower than those found in this study. Thedifferences in cost estimates are likely due to dif-ferences in sample populations. The mean age intheir sample was 65 years, and hip and pelvicfractures accounted for 31% of the fracture types,whereas these fracture types accounted for only21% of the fractures in the Medicare sample inthis study.

These results are also consistent with Euro-pean studies. Based on results from the SOTI(Spinal Osteoporosis Therapeutic Intervention)and TROPOS (the Treatment of Peripheral Osteo-porosis Study) clinical trials in five Europeancountries, Bouee et al.[24] found that hip fractureswere associated with the highest costs, followedby leg, humerus and pelvis fractures, all of whichhad similar costs. Levy et al.[23] found a similarpattern in the relative costs of NV fractures inmen in France. Bouee et al.[24] found a range ofcost estimates for NV fractures, reporting directcosts by fracture type and by country in h, year2002 values. Their estimates for total costs rangefrom a low of approximately h565 (approxi-mately $US560 in 2006 values) for rib fractures inSpain to a high of approximately h9907 (ap-proximately $US9900 in 2006 values) for hipfractures in France. That these European esti-mates were lower than US results may be due inpart to differences in demographics and costs ofmedical services.

Because of the nature of the data and thecondition studied, certain limitations apply. Inparticular, relying on claims data raises the pos-sibility of miscodings and misdiagnoses, and doesnot provide complete clinical detail. First, pre-vious studies have noted that using ICD-9-CMcodes to identify osteoporosis tends to under-estimate the true prevalence of the condition in apopulation, particularly among patients withmultiple co-morbid conditions.[35,36] To the ex-tent that such underestimation is present in thisstudy and resulted in patients with greater num-bers of co-morbidities being excluded from thesamples, the costs and resource use results mayhave been biased downwards. In addition, medi-cal claims with diagnoses for NV fractures do notindicate whether they are incident fractures. De-spite imposing a 6-month washout to increase the

likelihood that only incident fractures wereidentified, it is possible that a portion of thefracture patients in this study may have beenfollow-up patients or patients who received ima-ging for a suspected fracture but did not actuallyhave one. Also, this study assessed only incre-mental costs over the year following an NVfracture. Given the substantial incremental costsand the long periods of recovery and rehabilita-tion typically associated with NV fractures, theremay be additional longer term (i.e. 12–24 monthspost-fracture) incremental costs, which futureresearch may wish to examine. For all these rea-sons, this study may understate the full economicburden of NV fracture patients.

Additionally, Medicare results were limited bythe lack of data on prescription drugs. For theprivately insured sample, incremental drug costsaccounted for 13% of incremental direct costsamong all patients, and 9% of incremental totalcosts among employees only. By comparison,imputed incremental drug costs accounted foronly 3% of Medicare incremental direct costs. Inabsolute terms, imputed Medicare incrementaldrug costs ($US396) were also lower than drugcosts for privately insured patients ($US779).These findings suggest that the imputed drugcosts may understate the true costs of prescrip-tion drugs for NV fracture Medicare patients.The generalized linear regression model used hereto estimate direct costs has not been validatedand additional testing could be useful. As Medi-care Part D data become available, research ex-amining the drug costs of Medicare NV fracturepatients as well as the impact of the ‘doughnuthole’ would be informative.

Lastly, this study focused exclusively on pa-tients with clinically diagnosed osteoporosis. Inaddition to patients with osteoporosis who re-main undiagnosed, there are other patients whoare affected by NV fractures. For example, theNational Osteoporosis Risk Assessment (NORA)study reported that the total incidence of frac-tures in the osteopenic (i.e. low bone mass, butnot so low as to be classified as osteoporosis)population exceeds the incidence in the osteo-porosis population.[11] Further research thatincludes patients with osteopenia or other risk



factors for fractures but without confirmedosteoporosis would more fully capture the totalcost burden of NV fractures.

Conclusion

Due to their weakened bone structure, patientswith osteoporosis have increased risk of frac-tures. These fractures impose significant burdenson patients, requiring medical treatment and re-ducing the patient’s functional ability at homeand at work. Findings reported here are con-sistent with past research that has found thatosteoporotic patients with NV fractures impose asubstantial economic burden on their healthcarepayers.[6,7,34] Among the employed populationwith osteoporosis and NV fractures, there is asubstantial economic burden associated with lostproductivity in the workplace in addition to thealready large healthcare cost burden. While hipfracture patients are more costly on a per patientbasis, NVNH fracture patients are associatedwith a larger percentage of fractures and a greaterproportion of aggregate incremental costs forpatients with both private insurance and Medi-care. Therefore, better identification of patientsat risk and appropriate treatment using therapiesproven effective in reducing fractures acrossmultiple skeletal sites could lead to reductions inthe economic burden of NV fractures.

Acknowledgements

This research project was funded by Eli Lilly and Com-pany. Drs Russel Burge and Eric Edgell are employees of EliLilly and Company in the Global Health Outcomes depart-ment. Eli Lilly and Company markets pharmacologicaltherapies for the prevention and treatment of osteoporosis.

The authors thank Susan Dennett of Eli Lilly and Com-pany for her contributions to the study design. The authorsalso thank Anna Gu for providing computer programminginput on various aspects of the analysis.

References1. The World Health Organization Scientific Group. Preven-

tion and management of osteoporosis [technical reportseries no. 921]. Geneva, Switzerland: WHO, 2003

2. National Osteoporosis Foundation. Osteoporosis fastfacts [online]. Available from URL: http://www.nof.org/osteoporosis/diseasefacts.htm [Accessed 2009 Mar 6]

3. National Institute of Arthritis andMusculoskeletal and SkinDiseases. Fact sheets: osteoporosis overview [online].Available from URL: http://www.niams.nih.gov/Health_Info/Bone/Osteoporosis/default.asp [Accessed 2009Mar 6]

4. Pasco JA, Sanders KM,Hoekstra FM, et al. The human costof fracture. Osteoporos Int 2005 Dec; 16 (12): 2046-52

5. Srivastava M, Deal C. Osteoporosis in elderly: preventionand treatment. Clin Geriatr Med 2002; 18: 529-55

6. Burge R, Dawson-Hughes B, Solomon DH, et al. Incidenceand economic burden of osteoporosis-related fractures in theUnited States, 2005–2025. J BoneMinerRes 2007; 22: 465-75

7. Tosteson AN, Gabriel SE, Grove MR, et al. Impact of hipand vertebral fractures on quality-adjusted life years.Osteoporos Int 2001; 12: 1042-9

8. Boonen S, Singer AJ. Osteoporosis management: impactof fracture type on cost and quality of life in patientsat risk for fracture I. Curr Med Res Opin 2008; 24 (6):1781-8

9. Melton LJ, Gabriel SE, Crowson CS, et al. Cost-equivalenceof different osteoporotic fractures. Osteoporos Int 2003;14: 383-8

10. Orsini LS, Rousculp MD, Long SR, et al. Health care utili-zation and expenditures in the United States: a study of os-teoporosis-related fractures. Osteoporos Int 2005; 16: 359-71

11. Ray NF, Chan JK, Thamer M, et al. Medical expendituresfor the treatment of osteoporotic fractures in the UnitedStates in 1995: report from the National OsteoporosisFoundation. J Bone Miner Res 1997; 10 (1): 24-35

12. Melton LJ. Adverse outcomes of osteoporotic fractures inthe general population. J Bone Miner Res 2003 Jun; 18 (6):1139-41

13. Ohlsfeldt RL, Borisov NN, Steer RL. Fragility fracture-related direct medical costs in the first year following anonvertebral fracture in a managed care setting. Osteo-porosis Int 2006; 17: 252-8

14. Centers for Medicare and Medicaid Services. Medicare en-rollment – aged beneficiaries: as of July 2006 [online].Available from URL: http://www.cms.hhs.gov/MedicareEnRpts/Downloads/06Aged.pdf [Accessed 2009 Mar 9]

15. National Economic Council/Domestic Policy Council.America’s seniors and Medicare: challenges for today andtomorrow [online]. Available from URL: http://clinton5.nara.gov/media/pdf/med_exec_summ_2000.pdf [Accessed2009 Mar 9]

16. Purcell P. Older workers: employment and retirementtrends. CRS report for Congress, updated September 7,2007 [online]. Available fromURL: http://opencrs.cdt.org/rpts/RL30629_20070907.pdf [Accessed 2009 Mar 9]

17. Owen RA, Melton 3rd LJ, Gallagher JC, et al. The nationalcost of acute care of hip fractures associated with osteo-porosis. Clin Orthop Relat Res 1980 Jul-Aug; 150: 172-6

18. Martin BC, Chisholm MA, Kotzan JA. Isolating the cost ofosteoporosis-related fracture for postmenopausal women:a population-based study. Gerontology 2001 Jan-Feb; 47(1): 21-9

19. Lindsay R, Burge RT, Strauss M. One year outcomes andcosts following a vertebral fracture. Osteoporos Int 2005;16: 78-85

20. Chrischilles E, ShiremanT,WallaceR. Costs and health effectsof osteoporotic fractures. Bone 1994 Jul-Aug; 15 (4): 377-86

408 Pike et al.


21. King AB, Saag KG, Burge RT, et al. Fracture Reduc-tion Affects Medicare Economics (FRAME): impact ofincreased osteoporosis diagnosis and treatment. Osteo-poros Int 2005; 16: 1545-57

22. Zethraeus N, Borgstrom F, Johnell O, et al. Costs andquality of life associated with osteoporosis related frac-tures: results from a Swedish survey [Working Paper Seriesin Economics and Finance 2002, no. 512; online]. Availablefrom URL: http://swopec.hhs.se/hastef/abs/hastef0512.htm[Accessed 2009 Sep 10]

23. Levy P, Levy E, AudranM, et al. The cost of osteoporosis inmen: the French situation. Bone 2002 Apr; 30 (4): 631-6

24. Bouee S, Lafuma A, Fagnani F, et al. Estimation of directunit costs associated with non-vertebral osteoporotic frac-tures in five European countries. Rheumatol Int 2006 Oct;26 (12):1063-72

25. Chen Z, Maricic M, Aragaki AK, et al. Fracture risk in-creases after diagnosis of breast or other cancers in post-menopausal women: results from the Women’s HealthInitiative. Osteoporos Int 2009 Apr; 20 (4): 527-36

26. Abelson A. A review of Paget’s disease of bone with a focuson the efficacy and safety of zoledronic acid 5mg. CurrMed Res Opin 2008 Mar; 24 (3): 695-705

27. Blouin S, Libouban H, Moreau MF, et al. Orchidectomymodels of osteoporosis.MethodsMol Bio 2008; 455: 125-34

28. Hasselman CT, Vogt MT, Stone KL, et al. Foot and anklefractures in elderly white women: incidence and risk fac-tors. J Bone Joint Surg Am 2003 May; 85-A (5): 820-4

29. Kaye RA. Insufficiency stress fractures of the foot and anklein postmenopausal women. Foot Ankle Int 1998 Apr; 19(4): 221-4

30. Siris ES, Miller PD, Barrett-Connor E, et al. Identificationand fracture outcomes of undiagnosed low bone mineraldensity in postmenopausal women: results from theNationalOsteoporosis Risk Assessment. JAMA 2001; 286: 2815-22

31. Charlson ME, Pompei P, Ales KL, et al. A new method ofclassifying prognostic comorbidity in longitudinal studies:development and validation. J Chronic Dis 1987; 40 (5):373-83

32. Romano PS, Roos LL, Jollis J. Adapting a clinical co-morbidity index for use with ICD-9-CM administrativedata. J Clin Epidemiol 1993; 46 (10): 1075-9

33. Akobundu E, Ju J, Blatt L, et al. Cost-of-illness studies:a review of current methods. Pharmacoeconomics 2006;24 (9): 869-90

34. RousculpMD, Long SR,Wang S, et al. Economic burden ofosteoporosis-related fractures in Medicaid. Value Health2007; 10 (2): 144-52

35. Gunter MJ, Beaton SJ, Brenneman SK, et al. Managementof osteoporosis in women aged 50 and older with osteo-porosis-related fractures in a managed care population.Dis Manag 2003 Summer; 6 (2): 83-91

36. Sasser AC, Rousculp MD, Birnbaum HG, et al. Economicburden of osteoporosis, breast cancer, and cardiovasculardisease among postmenopausal women in an employedpopulation. Womens Health Issues 2005 May-Jun; 15 (3):97-108

Correspondence: Dr Howard G. Birnbaum, 111 HuntingtonAve, 10th Floor, Boston, MA 02199, USA.E-mail: [email protected]



Direct and Indirect Costs of Non-Vertebral Fracture Patients with Osteoporosis in the US

Documents

Transcript of Direct and Indirect Costs of Non-Vertebral Fracture Patients with Osteoporosis in the US