The direct and indirect costs of long bone fractures in a working age US population

Abstract

Objective:

Information regarding the burden of fractures is limited, especially among working age patients. The objective of this study was to evaluate the direct and indirect costs associated with long bone fractures in a working age population using real-world claims data.

Methods:

This was a claims-based retrospective analysis, comparing adult patients in the 6 months before and 6 months after a long bone fracture between 1/1/2001 and 12/31/2008 using the MarketScan Research Databases. Outcomes included direct medical costs and utilization, as well as work absenteeism and short term disability, which was available for a sub-set of the patients. Observed and adjusted incremental costs (i.e., the difference in costs before and after a fracture) were evaluated and reported in 2008 US$.

Results:

A total of 208,094 patients with at least one fracture were included in the study. Six, mutually exclusive fracture cohorts were evaluated: tibia shaft (n = 49,839), radius (n = 97,585), hip (n = 11,585), femur (n = 6788), humerus (n = 29,884), and those with multiple long bone fractures (n = 12,413). Average unadjusted direct costs in the 6-months before a long bone fracture ranged from $3291 (radius) to $12,923 (hip). The average incremental direct cost increase in the 6-months following a fracture ranged from $5707 (radius) to $39,041 (multiple fractures). Incremental absenteeism costs ranged from $950 (radius) to $2600 (multiple fractures), while incremental short-term disability costs ranged from $2050 (radius) to $4600 (multiple fractures).

Conclusions:

The results of this study indicate that long bone fractures are costly, both in terms of direct medical costs and lost productivity. Workplace absences and short-term disability represent a significant component of the burden of long bone fractures. These results may not be generalizable to all patients with fractures in the US, and do not reflect the burden of undiagnosed or sub-clinical fractures.

Keywords::

• Fractures
• Administrative claims data
• Lost productivity
• Burden of illness

Introduction

Musculoskeletal injuries are common, with an estimated 50 million injuries each year in the US requiring formal medical treatment1. Lower-extremity injuries, which account for 31% of all body injuries2, and distal radial fractures, which comprise one-sixth of all fractures seen in the emergency room3, are the most frequently occurring of such injuries. Given that increasing age and the presence of osteoporosis are the two leading drivers of fractures, much of the burden of fracture literature focuses on osteoporotic fractures among the elderly, and thus estimates of the prevalence and burden of fracture among the working age population are difficult to find. A recent study of adults aged 50 and over, however, suggests that men and women aged 50–64 account for ∼ 30% of all hip, vertebral, and non-hip, non-vertebral fractures4. In addition, most tibia fractures occur in working-age individuals under the age of 505. Therefore, the potential burden of not only medical costs associated with fracture repair but also the related lost work productivity and disability to the working age population and their employers could be substantial.

Both medical and productivity loss costs are likely to vary according to healing and rehabilitation time. Although healing time varies according to the location of the fracture, a study by Parziale6 indicates a range of anywhere between 6–8 weeks for a proximal humerus fracture and up to 12–16 weeks for a fracture of the femur. Full rehabilitation can take an additional 12–52 and 15–30 weeks for humerus and femur fractures, respectively. One prospective study of 158 adults aged 18–64 with a unilateral lower extremity fracture found that among those working prior to a fracture, 28% did not return to work within 12 months of the fracture7. Furthermore, at 30 months post-fracture, 17% had a mild disability, 12% had a moderate disability, and 7% had a severe disability as evaluated by the Sickness Impact Profile (SIP). Another Canadian study reported an average of 9.2 weeks of lost work for distal radial fractures; 21% reported no missed work time, which suggests a wide variation in recovery8.

Pike et al.9 examined both the medical and lost productivity costs in a sample of privately insured individuals with osteoporosis and a non-vertebral fracture. Using a retrospective claims data analysis, they estimated that the average annual incremental medical costs across all types of non-vertebral fractures were $5961. Among their sub-sample with lost productivity data, incremental absenteeism and short-term disability costs averaged roughly 2 weeks at a cost of $1956 and accounted for almost 30% of the total direct and indirect costs of the fracture.

In an effort to identify the costs of osteoporotic fractures, both the Shi et al.4 and Pike et al.9 studies excluded fractures that were likely the result of trauma. Thus, the results from these studies are not generalizable to the broader population of fracture patients, including those with complicated fractures and those without osteoporosis. In addition, only a handful of studies have examined the working age population. This study was designed to estimate both the direct costs of healthcare utilization and the indirect costs of lost work productivity associated with different types of long bone fractures among an under-65 population. We defined long bone fractures as fractures of the tibia shaft, radius, femur, humerus, or hip.

Methods

Data source

A claims-based retrospective analysis was undertaken that compared the medical and productivity costs before and after a long bone fracture between January 1, 2001 and December 31, 2008 using the two Truven Health Analytics MarketScan® Research Databases that contain data from large employers and managed care organizations: the Commercial Database and the Health and Productivity Management (HPM) Database (Ann Arbor, MI). The claims in the Commercial Database provide detailed cost and utilization data for healthcare services performed in both inpatient and outpatient settings. These covered lives comprise individuals from ∼ 90 large employers and health plans across the US. Insurance coverage for these individuals is provided under a variety of fee-for-service and capitated health plans, including exclusive provider organizations (EPO), preferred provider organizations (PPO), point of service plans (POS), indemnity plans, and health maintenance organizations (HMO). The HPM Database includes detailed information on workplace absences, short-term disability (STD), and worker’s compensation for a sub-set of individuals in the larger Commercial Database. The productivity and medical claims data are linked to outpatient prescription drug claims and person-level enrollment data through the use of unique enrollee identifiers.

Patient selection

All patients with a long bone fracture between January 1, 2001 and December 31, 2008 were identified and grouped into one of six mutually exclusive categories based on the following diagnosis codes: (1) tibia shaft fracture (ICD-9 codes 823.xx, excluding 823.4x), (2) distal radius fracture (ICD-9 codes 813.xx), (3) hip fracture (ICD-9 codes 820.xx), (4) femur fracture (ICD-9 codes 821.xx), (5) humerus fracture (ICD-9 codes 812.xx), and (6) those with multiple long bone fractures, as defined by more than one fracture within 7 days. These definitions were adapted from previous work in the area4.

To be included in the analysis sample, patients had to have had at least one non-diagnostic claim (i.e., not an x-ray claim) with either a primary or secondary diagnosis for a fracture of interest, and a minimum of 6 months continuous enrollment with concurrent pharmacy coverage prior to and following the index fracture date.

Patients were excluded if they were less than 18 years of age at the index event, were hospitalized at index and were not discharged within 30 days, had any type of fracture during the 6-month pre-period, or who were pregnant at any time during the 6-month pre-period or 6-month post-period.

The six main fracture cohorts were mutually exclusive, meaning that patients were assigned to a fracture cohort based on their first eligible long bone fracture type, starting on January 1, 2001 and ending on December 31, 2008. Patients could only appear in one fracture cohort.

Outcomes

Three types of outcomes were evaluated: (1) the direct medical costs, (2) healthcare resource utilization, and (3) the indirect work productivity costs associated with work days lost due to absenteeism, and short-term disability (STD).

Direct healthcare cost was defined as the total payment (by both patient and payer) from paid adjudicated claims in the Commercial Database. Costs were reported for the 6-month pre- and post-periods, and were adjusted to 2008 US dollars using the Medical Care Consumer Price Index, as provided by the US Bureau of Labor Statistics10. Costs for services provided under capitation insurance coverage were estimated from encounter records using fee-for-service equivalents defined by procedure code and region. The incremental cost due to fracture was estimated individually for each patient by subtracting their pre-period all-cause medical costs from their post-period all-cause medical costs; thus, each patient served as his or her own control. Medical costs included inpatient and outpatient services, emergency room visits, and pharmaceuticals. Average incremental resource utilization within each of the service types was also calculated by taking the difference between the post-period and pre-period utilization of inpatient, outpatient, emergency room services, and pharmaceuticals. These services were further broken out according to those claims having a fracture-related primary or secondary diagnosis and those without a fracture-related diagnosis. This approach was also used for short-term disability and absenteeism costs.

Costs occurring on or after the index date were included as post-index period costs. If a patient’s index date occurred during an inpatient stay, then the entire inpatient stay was included in the post-index period. This effectively reset the index date to the beginning of the inpatient stay containing the index event. The reset index date was used to differentiate pre- and post-period indirect, outpatient, and prescription drug costs as well. Patients who had a fracture during an inpatient stay were qualitatively examined to attempt to determine if the fracture occurred in the hospital and were flagged for possible exclusion.

Indirect cost and utilization variables were estimated for the sub-sets of patients who were primary coverage holder employees and had eligibility in the HPM Database for absences or STD during the calendar years covering their study period. Patients were identified by whether they had absence eligibility and whether they had any time lost due to absenteeism. In addition, total days lost were averaged across all patients with absence eligibility as well as for the sub-set of patients who had any time lost due to absenteeism. The cost of absenteeism was calculated as the number of hours of reported absenteeism multiplied by $30 (an average rate per hour)11–13. This was provided both for all patients with absence eligibility and for the sub-set of patients who had any time lost due to absenteeism.

As with absenteeism, patients were identified by whether they had STD eligibility and whether they had any time lost due to STD. Total days lost were averaged for all patients with STD eligibility and for the sub-set of patients who had any days lost due to STD. This approach was used for both pre- and post-fracture lost productivity. Applying the same $30 per hour salary for absenteeism costs, and under the assumption that the general practice of employers covering 70% of salary during STD holds, the cost of each STD lost hour was estimated to be $21. Average costs of STD were calculated both for all patients with STD eligibility and for the sub-set of patients who had any days lost due to STD. Patients within the Commercial database can have absenteeism data eligibility, STD data eligibility, both, or neither. Lost work time due to absenteeism and STD data are exclusive of each other.

Explanatory variables

Patients’ age, gender, insurance plan type (Comprehensive, HMO, POS, PPO, and other), and geographic region (Northeast, North Central, South, West, and Unknown) were identified at baseline. Pre-period clinical characteristics included the presence of comorbidities known to be predictors of fracture such as osteoporosis, rheumatoid arthritis, diabetes and other endocrine diseases, liver disease, thyroid disease, disorders of the bone and cartilage, and cancer. Co-morbid conditions were identified by ICD-9 codes for each of the conditions listed as a primary or secondary diagnosis on any claim during the 6-month pre-period. Prescription claims for medications known to be associated with fractures were also flagged during the 6-month pre-period, including medications to treat osteoporosis, glucocorticoids (excluding non-systemic forms such as topical or inhaled applications), anticonvulsants, immunosuppressants, thiazolidinediones, hormone deprivation therapy, proton pump inhibitors, and NSAIDs or COX-2 inhibitors. In addition, the Charlson Comorbidity Index (CCI) was calculated based on ICD-9 codes in 17 categories of chronic conditions that appeared on claims during the pre-period14,15. While most of the 17 categories have a weight of 1, four categories have a weight of 2, one (moderate or severe liver disease) has a weight of 3, and two (metastatic cancer and HIV) have weights of 6.

Characteristics of the index fracture were captured during the 6-month post-period. Each fracture was flagged as being open (ICD-9 codes with a suffix of .1x, .3x, .5x, and .9x) or closed (ICD-9 codes with a suffix of .0x, .2x, and .8x). Poor healing outcomes were identified during the period from the 31st day after the index date to the end of follow-up and included claims with diagnosis code 733.81 (non-union of fracture), or 733.82 (mal-union of fracture), or any of the procedure codes associated with repair of a non-union or mal-union fracture. In addition, new fractures occurring in the post-period were captured for the five long bone cohort-defining fractures and vertebral fractures (ICD-9 diagnosis code 805.xx). Unadjusted differences in costs prior to and following fracture were compared using paired t-tests.

Multivariate analysis

Multivariate analysis was used to adjust for differences associated with the demographic and clinical factors. First, generalized linear models (GLM) were implemented on the pooled sample of all patients (all six fracture cohorts combined), in which the dependent variables were the total post-period expenditure and the difference in expenditure between pre- and post-periods. The key independent variables were dummy indicators of fracture type representing the six different fracture cohorts. The control variables included: internal fixation; age; gender; insurance type; region; urbanicity; industry; relationship to employee; employment status (hourly, salaried, unknown); union status (union, non-union, unknown); pre-period total healthcare cost; CCI score; and binary indicator variables for the presence of pre-period co-morbid clinical conditions and usage of medications that may impact the likelihood of fracture. The distributions of the outcome variables were examined and Park tests were conducted16,17. Log link and gamma variance functions were chosen for the outcome of total post-period cost. Identity link and Gaussian variance functions were used for the outcome of pre–post difference in cost. Several variance functions were tested (gaussian, gamma, negative binomial) and the AIC was tested to determine the model with the best fit.

The outcomes of post-period absenteeism and STD costs and the pre–post difference (change) in those costs (6-month post-period costs minus 6-month pre-period costs) were also examined using the pooled sample that combined all six facture cohorts. GLM regressions with log link and gamma variance functions were applied to post-period absenteeism costs. A two-part model was implemented on the STD cost in the post-period, in which the first part was a logistic regression of any STD cost, and the second part was a GLM regression of STD cost among patients with positive STD cost, with log link and gamma variance. The outcomes of pre–post difference in absenteeism and STD costs were fitted with GLMs with identity link and Gaussian variance functions. The dummy indicators of fracture types were included as key independent variables to capture the difference in outcomes across types of fracture. The same set of control variables were included as in the cost models, except where small samples required that certain covariates be dropped. The impact of internal fixation on fracture costs was also assessed using GLM regressions. For these models, the key independent variable was a binary indicator of internal fixation within 7 days of the index fracture diagnosis and the control variables were the same as the covariates mentioned above.

Adjusted post-period costs or pre–post difference in costs were predicted for each individual based on the regression estimates, assuming that everyone in the dataset had a tibia fracture, radius fracture, hip fracture, femur fracture, humerus fracture, or multiple fractures. The means for the predicted costs were reported. The means of adjusted outcomes for patients with and without internal fixation were also reported by each type of fracture. At no point were fracture cohorts compared to each other. All analyses were conducted using Stata/MP version 1118.

Results

A total of 208,094 patients were included in the study. Fractures of the radius were most common (n = 97,585), followed by the tibia (n = 49,839), humerus (n = 29,884), multiple (n = 12,412), hip (n = 11,585), and femur (n = 6788). Demographic characteristics for each of the fracture groups are presented in Table 1. Females appeared to comprise a larger percentage of each fracture cohort than did males, with the exception of femur fracture (51.9% male, 48.1% female). The humerus cohort had the highest proportion of females (62.0% female), followed by radius fracture (60.2% female), and hip fracture (57.0% female). Mean age ranged from a low of 43.4 years for femur fracture to a high of 52.1 years for hip fracture. The pattern of results observed for geographic region and insurance plan type was reflective of the overall nature of the employers and health plans contributing data to the Commercial Database over the measurement time period. The most common type of health plan was PPO (52.6–54.3%), followed by HMO (16.6–18.9%), and comprehensive (10.4–15.3%).

Table 1.  Baseline demographic characteristics of fracture cohorts.

	Tibia (n = 49,839)	Radius (n = 97,585)	Hip (n = 11,585)	Femur (n = 6788)	Humerus (n = 29,884)	Multiple (n = 12,413)
Female (%)	52.1%	60.2%	57.0%	48.1%	62.0%	53.2%
Age: Mean (SD)	45.4 (12.7)	45.8 (13.1)	52.1 (11.1)	43.4 (14.2)	49.0 (12.1)	46.9 (13.1)
Geographic Region
Northeast	10.6%	10.5%	11.7%	9.5%	11.2%	9.9%
North Central	26.6%	27.1%	26.7%	26.6%	25.7%	25.0%
South	42.1%	42.1%	42.5%	44.6%	41.4%	45.9%
West	20.3%	19.8%	18.5%	18.7%	21.2%	18.9%
Unknown	0.5%	0.5%	0.5%	0.5%	0.5%	0.3%
Insurance Plan Type
Comprehensive	10.4%	10.7%	15.3%	12.2%	11.6%	10.9%
HMO	18.9%	18.7%	16.6%	18.8%	18.8%	18.3%
POS	10.9%	10.9%	9.8%	10.2%	10.5%	10.4%
PPO	53.2%	53.3%	52.6%	53.2%	52.8%	54.3%
Other	6.5%	6.4%	5.7%	5.7%	6.3%	6.1%
Baseline Clinical Characteristics
CCI: Mean (SD)	0.27 (0.84)	0.23 (0.74)	0.74 (1.61)	0.47 (1.33)	0.41 (1.12)	0.41 (1.12)
Comorbid Conditions^a
Osteoporosis	0.9%	1.1%	2.5%	1.4%	1.3%	1.2%
Diabetes	7.4%	5.7%	11.5%	8.1%	10.4%	9.2%
Rheumatoid Arthritis	0.8%	0.7%	2.7%	1.7%	1.4%	1.4%
Endocrine Disease (excluding diabetes)	1.0%	1.0%	1.9%	1.3%	1.2%	1.3%
Liver Disease	0.9%	0.9%	2.2%	1.2%	1.3%	1.3%
Breast Cancer	0.9%	1.3%	2.5%	1.5%	1.6%	1.6%
Bone Cancer	0.1%	0.1%	1.5%	1.7%	0.5%	0.5%
Other Cancer^b	1.9%	2.0%	4.7%	2.7%	2.6%	2.7%
Thyroid Disease	3.3%	3.5%	4.7%	3.3%	3.7%	3.3%
Unspec. disorder of bone and cartilage	1.0%	1.1%	2.4%	2.3%	1.4%	0.9%
Concomitant Medications
Osteoporosis	9.8%	11.3%	17.0%	9.9%	12.9%	11.1%
Glucocorticoids	14.2%	13.2%	22.7%	18.0%	15.8%	15.2%
Anticonvulsants	6.8%	6.0%	13.0%	8.3%	8.5%	8.3%
Immunosuppressants	2.6%	2.4%	6.8%	4.0%	3.8%	3.9%
Thiazolidinediones (TZDs)	3.0%	2.3%	3.6%	2.8%	4.7%	3.6%
Hormone deprivation therapy	0.6%	0.8%	1.4%	0.9%	1.0%	0.9%
Proton Pump Inhibitors (PPIs)	11.9%	11.1%	18.2%	13.1%	14.5%	13.0%
NSAIDs/Cox-2 Inhibitors	34.2%	29.6%	49.0%	44.9%	37.2%	30.3%

^aConditions not shown due to prevalence < 1% across all groups are: HIV, alcoholism, osteodystrophy, nephritis, Paget’s disease, osteomyelitis.

^bExcluding bone cancer, breast cancer, and melanoma.

The pre-period CCI suggests that, overall, the patient population had few chronic conditions prior to fracture, with CCI’s ranging from 0.23–0.47 for all but the hip fracture cohort, which had an average score of 0.74. Of all the conditions associated with fracture, only diabetes and thyroid disease had a consistent prevalence greater than 3% across the cohorts. While the diagnosed prevalence of osteoporosis was less than 2% for all but the hip fracture cohort, the use of osteoporosis medications ranged from 9.8% (tibia) to 17.0% among hip fracture patients. The use of glucocorticoids, anticonvulsants, PPIs, and NSAIDs were also common in each of the fracture cohorts, with usage ranging between 6–13% for anticonvulsants to between 29.6–49.0% for NSAIDs (Table 1). Overall, however, the data suggest that this is a relatively healthy population at baseline (i.e., pre-fracture).

Characteristics of the fractures varied across the cohorts (Table 2). The rates of open fracture, for example, ranged from 2.5% for humerus fracture to 8.0% for multiple long bone fracture. Similar variations were noted for poor fracture healing and experiencing another fracture. Substantial differences in the rate of having a fracture that required internal fixation were also observed, with just under 16% of radius fractures requiring internal fixation to 56% of multiple fractures.

Table 2.  Fracture characteristics of cohorts.

	Tibia (n = 49,839)	Radius (n = 97,585)	Hip (n = 11,585)	Femur (n = 6,788)	Humerus (n = 29,884)	Multiple (n = 12,413)
Fracture profile/characteristics
Open fracture^a	7.3%	2.7%	5.4%	7.8%	2.5%	8.0%
Poor fracture healing outcome^b	2.8%	1.6%	1.7%	3.9%	2.7%	5.9%
Fractures during 6 month post-period
Other type of long bone fracture^c	1.3%	1.9%	7.1%	8.9%	4.1%	5.0%
Days since index (mean, SD)	60.7 (51.6)	51.7 (44.7)	51.8 (46.9)	46.9 (43.4)	49.2 (46.9)	46.0 (40.4)
Vertebral fractured	1.0%	1.0%	2.6%	2.5%	1.3%	4.9%
Days since index (mean, SD)	13.5 (35.9)	17.8 (39.4)	19.5 (38.9)	12.2 (33.3)	19.8 (41.2)	6.8 (23.9)
Other type of long bone^c or vertebral fracture^d	2.3%	2.7%	9.3%	10.8%	5.2%	9.0%
Days since index (mean, SD)	39.6 (50.9)	40.0 (46.0)	43.0 (47.4)	39.1 (44.1)	41.8 (47.3)	24.9 (38.2)
Internal fixation (%)	23.7%	15.7%	36.0%	28.1%	10.3%	56.0%

^aBased on index long-bone fracture diagnosis.

^bNon-union or mal-union diagnosis or procedure for treating non-union or mal-union from day 31 after index date to end of post-period.

^cFrom 8 days after index to end of 6-month follow-up period.

^dAnytime during 6-month follow-up period.

Unadjusted healthcare costs in the pre- and post-periods, as well as the change from pre- to post-fracture, are depicted in Figure 1 according to the fracture cohort. Mean all-cause costs for the 6-month post-period exceeded the mean all-cause costs for the 6-month pre-period in every cost category, and all of the differences between the pre-period and post-period costs shown in the figure were statistically significant (p < 0.0001 for all cost-categories in all six cohorts). The observed mean pre- to post-period change in 6-month total all-cause healthcare costs ranged from a mean increase of $5707 (2008 US$) among the radius fracture cohort to a mean increase of $39,041 among the multiple fractures cohort. As might be expected, there appeared to be variation among the cohorts not only in the amount of the increase in total healthcare costs occurring after fracture, but also in the service areas in which cost increases were concentrated. In general, the largest cost increases after fracture were observed in inpatient care and other outpatient services (emergency room excluded). Relatively smaller increases were present among emergency room and outpatient pharmacy costs.

Figure 1.  Unadjusted all-cause healthcare costs in the pre- and post-period, and pre–post change in cost. All pre–post comparisons were statistically significant (p < 0.001).

There were no statistically significant differences found between the percentages of patients who were absent at least 1 day (p > 0.05 for all six of the main fracture cohorts). These percentages were high in the pre-period and post-period for all cohorts, and reflected the fact that over a 6-month time period most employees missed some amount of work due to general absenteeism, whether or not they had experienced a long bone fracture. However, while the proportion of patients with an absenteeism claim was unchanged, there were significant increases in the amount of absenteeism time (p < 0.01 for all six fracture cohorts). The mean increase in absenteeism days ranged from 3.7 days for radius fracture to 13.9 days for multiple fractures. Multiplying by a wage constant of $30 per hour, this was equivalent to increased costs of $886–$3337 during the 6-month period following long bone fracture (see Figure 2 and Table 3).

Figure 2.  Unadjusted absenteeism and STD costs, by fracture type. All pre–post comparisons were statistically significant (p < 0.001).

Table 3.  Incremental lost work productivity in post-fracture period (post-period minus pre-period work loss)^a.

	Tibia		Radius		Hip		Femur		Humerus		Multiple
Incremental Absenteeism
Patients with absence eligibility (n)	813		1463		119		83		450		159
Patients with any days lost (n, %)	−1	−0.1%	13.0	0.9%	1	0.8%	1	1.2%	0	0.0%	2	1.3%
Days lost (among eligible patients)  (Mean, SD)	6.9	21.5	3.7	15.9	9.8	28.6	10.3	24.9	6.3	20.1	13.9	33.4
Days lost (among patients with any)  (Mean, SD)	8.4	n/a	4.3	n/a	12.4	n/a	11.4	n/a	7.8	n/a	15.9	n/a
Incremental STD
Patients with STD eligibility (n)	2514		4587		371		273		1282		515
Patients with any days lost (n, %)	734	29.2%	810	17.7%	102	27.5%	55	20.2%	295	23.0%	216	41.9%
Days lost (among eligible patients)  (Mean, SD)	23.1	40.1	10.8	27.5	24.6	42.6	20.2	39.7	15.6	33.9	36.8	47.5
Days lost (among patients with any)  (Mean, SD)	40.6	n/a	24.6	n/a	37.0	n/a	37.1	n/a	28.2	n/a	42.4	n/a

^aAll pre–post differences are statistically significant at p < 0.005. Positive values indicate an increase in absenteeism or STD in the post-period relative to the pre-period. Negative values indicate a decrease in absenteeism or STD in the post-period relative to the pre-period.

Statistically significant increases in all-cause lost work productivity due to STD, however, were observed for all six fracture cohorts (p < 0.0001 for all comparisons). For example, the percentage of radius fracture patients with one or more days lost from work due to STD increased from 7.8% in the 6-month pre-period to 25.5% in the 6-month post-period, and for multiple fracture patients this percentage increased from 13.0% to 55.0%. From the 6-month pre-period to the 6-month post-period the mean increase in days lost from work due to STD ranged from 10.8 days for radius fracture to 36.8 for multiple fractures. Multiplying by $21 per hour, STD costs for employers ranged from $1820–$6177 during the 6-month period following long bone fracture (see Figure 2).

All-cause costs changed for some of the fracture groups after multivariate adjustments for demographic characteristics and comorbidities, resulting in notable shifts in the all-cause costs across the cohorts (Table 4). Costs for radius and humerus fracture patients increased by 19% and 29%, respectively, to $6805 and $10,842, while adjusted costs for multiple fracture patients decreased by 21% to $30,867. Analysis of the impact of internal fixation on all-cause pre–post cost differences demonstrated that, for all fracture groups except hip fracture, patients with internal fixation had costs that were 4.7–5.7-times higher than for those without internal fixation (data not shown). Among hip fracture patients, internal fixation resulted in costs that were 2.7-times higher.

Table 4.  Multivariate adjusted pre–post differences in 6-month costs by fracture cohort.

	Total all-cause direct healthcare costs	STD costs	Absenteeism costs	Lost productivity costs as % of direct costs
Tibia fracture	$10,070	$3,596	$1,570	51%
Radius fracture	$6,805	$2,054	$953	44%
Hip fracture	$15,631	$3,640	$2,361	38%
Femur fracture	$17,655	$3,083	$2,421	31%
Humerus fracture	$10,842	$3,172	$1,696	45%
Multiple fracture	$30,867	$4,647	$2,654	24%

STD, Short-term disability.

Multivariate adjustment of the absenteeism and STD costs demonstrated similar changes, as did the adjustment of the all-cause healthcare costs (Table 4). Changes due to adjustment of demographic and clinical characteristics were strongest for humerus fracture patients, who saw an increase of 12% and 21% in absenteeism and STD costs, respectively. Multiple fracture patients, on the other hand, had decreases of 21% and 25%, respectively.

Sensitivity analyses of the impact of internal fixation on lost work productivity were possible for the tibia, radius, and humerus cohorts. Patients with internal fixation had absenteeism costs that were 2.0–2.8-times higher, and STD costs that were 2.4–5.1-times higher than patients without internal fixation.

Discussion

This analysis of over 200,000 long bone fracture patients represents the largest fracture study to date of a working age population, regardless of osteoporosis and trauma status. The patients in this study include those with simple fractures as well as those with more complicated fractures, as indicated by the percentage of poor healing (range of 1.6–5.9% across fracture cohorts) and internal fixation (range of 10.3–56% across fracture cohorts) patients.

The adjusted direct medical costs associated with the fractures ranged from ∼ $6800 for radius to $18,000 for femur fractures; adjusted direct medical costs were $31,000 for patients with multiple fractures. The fact that noticeable changes in total costs occurred following adjustment for baseline demographic and clinical characteristics suggests an interaction between underlying individual patient comorbidities and their post-fracture care. Future research is necessary to determine which characteristics lead to higher post-fracture costs and which are associated with lower costs and potentially quicker healing. In the post-fracture time period, internal fixation was found to be a significant driver of cost variations such that costs for patients with internal fixation ranged from 2.7 - (hip) to 5.7-times (tibia) higher than for patients who did not require internal fixation. Internal fixation is likely an indicator of fracture severity, requiring far more extensive healthcare utilization than external utilization.

Previous studies of the cost of fracture have generated somewhat similar cost estimates than this study, although a number of differences in study design must be noted. Pike et al.9, for example, estimated that the annual cost of hip fracture among adults aged 18–64 was $12,562, while Shi et al.4 found the annual cost of hip fracture among those aged 50–64 to be $26,545. This study, however, estimates the 6-month cost of hip fracture among those 18–64 years old to be ∼ $15,631. There are several reasons why these estimates may be different than those in previous studies. First, our study used a pre–post design whereby each patient is their own control, while the Pike et al. and Shi et al. studies used case-control designs. Because fractures are reasonably considered exogenous events, we believe that the own-control design has some advantages in addressing baseline cost and utilization differences. Second, both Pike et al. and Shi et al. attempted to identify only osteoporosis-related fractures by eliminating open fractures. The Pike et al. study also excluded fractures related to trauma. Third, while our study only examined 6-month costs, this does not imply that the annual costs would be double our estimate. It is likely that the majority of fracture-related costs occur within the first 6-months of the fracture. Lastly, some of the cost difference is explained by healthcare cost inflation, as the Pike et al. and Shi et al. studies used 2006 dollars, while our study uses 2008 dollars.

Estimated costs of productivity loss from absenteeism ranged from $950 for radius to $2600 for multiple fractures; estimated costs of productivity loss from STD ranged from $2050 for radius to $4600 for multiple fractures. In terms of actual work time lost, STD losses ranged from ∼ 11 days (or two working weeks) for radius to up to 37 days or 7 weeks for multiple fractures. Days lost due to absenteeism ranged from 4–14 days. As with healthcare costs, internal fixation increased STD time lost by 2.8-times for hip to 5.1-times for femur, and doubled absentee time.

These estimates of fracture-related work productivity loss are consistent with those reported by Pike et al.9, which estimated the mean cost of lost productivity to be ∼ $1956 for absenteeism and STD combined across all fracture types. Given that Pike et al. only report the combined mean for the two sources of productivity loss, it is difficult to compare our numbers directly with their numbers. Because our samples for absenteeism and STD do not have a perfect overlap, we refrain from providing a combined mean cost. However, keeping in mind that not all fracture patients work and not all who work experience significant lost time, it is possible that a combined estimate of work loss cost in our study would be similar to that found by Pike et al. This is an interesting result in that our population is generally younger and healthier. It appears that fracture costs are not merely a signal of a general decline in health, but are possibly a sudden, exogenous health shock that generates significant costs to the patient and the employer.

Limitations of this analysis include selection and information biases as well as general limitations of working with claims data. Because the MarketScan Research Databases rely on administrative claims data for clinical detail, the data are subject to data coding limitations and data entry error. This risk, however, should be consistent across patient cohorts. It is possible that diagnoses on claims were coded incorrectly or not coded at all, thereby potentially under-estimating the fracture-related resource utilization and costs. Race, socioeconomic status, anthropometric information, and mortality were unavailable in medical claims. In addition, only clinically diagnosed fractures were captured. Patients with sub-clinical fractures that did not reach medical attention and thus did not incur healthcare resource utilization or costs were not included in the study. While this study did capture lost productivity from absenteeism or short-term disability, it does not include presenteeism (i.e., the degree to which employees are present at work but not fully engaged or productive). Reduced presenteeism associated with fractures would add to the cost burden described in this analysis. Also, this study did not incorporate medical or administrative costs associated with worker’s compensation claims, potentially causing this study to under-estimate the burden of fractures to self-insured employers.

The MarketScan Commercial and HPM Databases include data provided largely by large US employers. Patients covered by health plans offered by small employers were under-represented. Medicare patients, uninsured patients, and patients covered by military health plans were excluded. Thus, findings from the study may not be generalizable to the whole US population or to fracture patients outside of the US. With regard to limitations of the study design, the 6-month requirement of continuous enrollment means that any fracture patient dying or leaving their employment within 6 months of their fracture was excluded. This may have resulted in study samples being healthier than the general population of patients with long bone fractures.

Conclusions

The results of this study demonstrate that fractures are common among a working aged population and are associated with both high direct costs and substantial productivity loss. For highly-paid individuals, employment costs from absenteeism caused by the fracture may easily exceed direct medical costs. Also, while over 95% of fractures heal normally, those who experience poor healing or internal fixation may have more than double the work losses and medical costs.

Transparency

Declaration of funding

This research was funded by Amgen, Inc.

Declaration of financial interests

TB is an employee of Amgen, Inc. and has received Amgen stock/stock options; MB and DE are employees of Truven Health Analytics (formerly Thomson Reuters Healthcare), which received a research contract to conduct this analysis.