Costs and quality of life associated with osteoporosis-related fractures in Sweden

Transcription

1 Osteoporos Int (2006) 17: DOI /s ORIGINAL ARTICLE Costs and quality of life associated with osteoporosis-related fractures in Sweden Fredrik Borgström. Niklas Zethraeus. Olof Johnell. Lars Lidgren. Sari Ponzer. Olle Svensson. Peter Abdon. Ewald Ornstein. Karl Lunsjö. Karl Göran Thorngren. Ingemar Sernbo. Clas Rehnberg. Bengt Jönsson Received: 27 June 2005 / Accepted: 19 September 2005 / Published online: 9 November 2005 # International Osteoporosis Foundation and National Osteoporosis Foundation 2005 Abstract This prospective observational data collection study assessed the cost and quality of life related to hip, vertebral and wrist fracture 1 year after the fracture, based on a patient sample consisting of 635 male and female F. Borgström. C. Rehnberg Medical Management Centre, Karolinska Institutet, Stockholm, Sweden F. Borgström (*) Stockholm Health Economics, Klarabergsgatan 33 3tr, Stockholm, Sweden fredrik.b@healtheconomics.se Tel.: Fax: N. Zethraeus. B. Jönsson Stockholm School of Economics, Stockholm, Sweden O. Johnell. I. Sernbo Department of Orthopaedics, Malmö General Hospital, Malmö, Sweden L. Lidgren. K. G. Thorngren Department of Orthopaedics, Lund University, Lund, Sweden S. Ponzer Department of Orthopaedics, Stockholm Söder Hospital, Karolinska Institutet, Stockholm, Sweden O. Svensson Division of Orthopedics, Umeå University Hospital, Umeå, Sweden P. Abdon Department of Orthopedics, Ystad Hospital, Ystad, Sweden E. Ornstein Department of Orthopedics, Hässleholm-Kristianstad Hospitals, Hässleholm, Sweden K. Lunsjö Department of Orthopaedics, Helsingborg Hospital, Helsingborg, Sweden patients surviving a year after fracture. Data regarding resource use and quality of life related to fractures was collected by questionnaires at baseline, 4 months and 12 months. Information was collected by the use of patients records, register sources and by asking the patient. Quality of life was estimated with the EQ-5D questionnaire. Costs were estimated from a societal perspective, including direct and indirect costs. The mean fracture-related cost the year after a hip, vertebral and wrist fracture were estimated, in euros ( ), at 14,221, 12,544 and 2,147, respectively [converted from Swedish krona (SEK) at an exchange rate of SEK/ ]. The mean reduction in quality of life was estimated at 0.17, 0.26 and 0.06 for hip, vertebral and wrist fracture, respectively. Based on the results, the yearly burden of osteoporosis in Sweden could be estimated at 0.5 billion (SEK 4.6 billion). The patient sample for vertebral fracture was fairly small and included a high proportion of fractures leading to hospitalization, but they indicate a higher cost and loss of quality of life related to vertebral fracture than previously perceived. Keywords Costs. Fracture. Osteoporosis. Quality of life Introduction Osteoporosis-related fractures are associated with a high degree of morbidity and mortality. In Sweden approximately 70,000 osteoporosis-related fractures occur annually [1]. The societal burden these fractures impose, both in terms of costs and in terms of reduction in survival and quality of life for the affected individuals, makes osteoporosis an important and increasing public health issue [1]. The risk of osteoporotic fractures can be reduced with different treatment strategies, such as behavioral changes, hip protectors and drugs. However, preventive measures cost money, and the cost-effectiveness of different intervention strategies must be documented. To assess the cost-effectiveness of different interventions it is important to have good estimates on costs and quality of life related to osteoporotic fractures that is potentially avoided by the treatment strategy.

2 638 There are studies that have previously estimated costs and/or the reduction in the quality of life related to osteoporotic fractures [2 16], but only a few of them are appropriate for use in economic evaluations. Not surprisingly, it is the hip fracture, the fracture type associated with the most severe mortality and morbidity, which has been most frequently investigated in these studies. This is mainly because patients with hip fracture are, in most cases, hospitalized, which generates the data available from hospital statistics. Consequences of other fracture types, particularly vertebral fractures, have, so far, been relatively poorly investigated. Many of the previous studies estimating the economic impact of fractures are not suitable for use in health economic evaluations, due to inappropriate data collection methodology, e.g., not all relevant cost items are included, resources are collected over too limited a time period, and the relevant costs that can be saved by preventing a fracture are not correctly estimated. With the purpose to obtain relevant estimates of costs and health-related quality of life consequences associated with osteoporotic fractures of the hip, vertebrae and wrist in Sweden, a prospective observational data collection study was launched in July The costs and effects of osteoporosis-related fractures study (the KOFOR-study) continuously enrols fracture patients at seven hospitals in Sweden and follows them over a period of 18 months after the fracture. Resource use is collected from a societal perspective, and the quality of life of patients after fracture is estimated with the EQ-5D questionnaire. The primary purpose of this article was to present the fracture-related costs and the impact on quality of life the first year after a sustained fracture, based on an interim analysis including 635 patients with either a hip, vertebral or a wrist fracture in the KOFOR study who survived the whole first year after the fracture. Other aims were to describe the impact of age, gender, previous fractures, and other factors that might have an impact on the fracture-related costs and impaired quality of life. Methods and materials When one is estimating the costs related to fractures it is important that only the extra cost incurred by the fracture is accounted for. The potential cost savings of avoiding a fracture could either be estimated by relating all costs the year before to all costs the year after fracture or by only including the costs relevant for the fracture. The first alternative demands large numbers of patients at risk of fracture, studied prospectively. An example of such a study is the Rotterdam study, where the incremental cost of hip and vertebral fracture was estimated [16]. If a retrospective study design is used, the selection of persons with a fracture is more practicable, since the fracture risk for a specific person is very small. This approach has also been used in previous Swedish fracture cost studies [8, 10], where costs before and after a fracture have been estimated based on register data. The main disadvantage with such studies is that it is hard to collect all relevant resource use. For example, it is rare for information about indirect costs and informal care to be available from any register. The advantage is that register studies can often provide fairly large sample sizes. In this study, fracturerelated resource use is collected prospectively after the fractures have occurred. The advantages of this approach is that it is possible to include all relevant cost items and that information prior the fracture does not have to be collected, which reduces the time needed for data collection. A disadvantage could be that, in some cases, it might be hard to determine whether a resource is related to the fracture or not. To our knowledge the KOFOR study is the first study using this data collection methodology. To test and validate the methodology the KOFOR study was preceded by a pilot study at Malmö University Hospital, in Sweden [17]. Study design and data collection The ongoing KOFOR study includes patients subjected to an osteoporosis-related fracture of the hip, vertebra, and wrist at the orthopedic departments at seven Swedish hospitals. 1 To be included, a patient has to be diagnosed for a fracture caused by low-energy trauma and be at least 50 years old. Vertebral fractures have to be confirmed by an X- ray examination. Patients seeking care for multiple fractures are not eligible for the study. Fractures caused by other diseases, such as cancer, are not included. Also, patients that are judged not to be able to complete the questionnaires due to dementia or other psychological problems are excluded from the study. To be eligible for the study the patient has to be included and interviewed within 4 weeks of the fracture event. Patients have to give their informed consent to be enrolled in the study prior to inclusion. The study was approved by local ethics committees. Data regarding resource use and quality of life related to fractures is collected by questionnaires at baseline, 4 months, 12 months and 18 months after fracture occurrence. The data collection questionnaire used was developed and tested at Malmö University Hospital [17]. At baseline patients characteristics, background information and perceived health status just before, and health status after, the occurrence of the fracture are collected. At the 4-month, 12-month and 18- month follow-up, information on resource use since last visit and current health status is collected. Patient-reported information at baseline is primarily collected via interviews in connection with the hospital visit related to the fracture event. If the interview cannot be conducted in relation to the hospital visit the information is collected via a telephone interview. Quality of life and patient-reported resource use at the follow-ups are collected via telephone interviews. 1 The participating centers were: Södersjukhuset in Stockholm, Malmö University Hospital, Lund University Hospital, Norrlands University Hospital, The Hospital in Ystad, Helsingborg Hospital and Hässleholm Hospital.

3 639 Resource use Information about use of resources was collected from patients records, register sources and by asking the patient. The resources were categorized into medical costs (hospitalization, outpatient care and pharmaceuticals), non-medical costs (community care and informal care) and indirect costs (i.e., loss of production). Community care consisted of special living arrangements, home care and transportation. Patient-reported resource use (community care, and informal care) and resources lost, i.e., indirect costs, were recorded for a shorter time during the study (1 month) to minimize recollection bias. months after fracture) the observed resource use during the past 4 weeks was extrapolated to the whole period. Informal care The patients was asked about the average number hours of fracture-related care given by relatives during the last 4 weeks prior to the interview at the 4-month and 12-month follow-ups. To obtain an estimate of the total informal care given in each period (0 4 months and 5 12 months after fracture) the observed resource use during the past 4 weeks was extrapolated over the whole period. Hospitalization For each episode of fracture-related hospitalization the number of bed days was recorded along with the ICD-10 main and other diagnoses and operation codes. Outpatient care Outpatient care consisted of outpatient surgery, physicians visits, nurses visits, physiotherapists and occupational therapists visits at the hospital, primary care visits at the office and at home, X-ray examinations, and telephone help. The resources were recorded for the whole period (i.e., 0 4 months after fracture and 4 12 months after fracture). Pharmaceuticals Data on the use of pharmaceuticals 2 deemed to be relevant for treatment of osteoporosis and fractures were collected. The patient was recorded as a user of a drug if she/he had, sometime during the period, been prescribed the pharmaceutical because of the fracture. If the patient had been a drug user then it was assumed that the drug had been taken during the whole period. Community care Each patient s living accommodation before fracture was recorded at baseline. At the 4-month and 12-month followups the patient was asked how many days she/he had stayed under any special living arrangements (e.g., nursing home and group living), due to the fracture, anytime during the last 4 weeks prior to the interview. The patients also stated the average number of hours of home help per week and the amount of transportation during the past 4 weeks that were related to the fracture. To obtain an estimate of the total resource use of community care (0 4 months and The following pharmaceutical categories were included: calcium and vitamin D, estrogens, bisphosphonates, glucocorticoids and pain medication. Loss of production At the baseline interview the patients gave information concerning their pre-fracture working status (full time work/part-time work/not employed). At the 4-month and 12-month follow-up interviews the patients stated their current working situation and, if they were working, how many days they had been on sick leave because of the fracture during the past 4 weeks. The stated sick leave was assumed to be the average monthly sick leave for the whole period (i.e., 0 4 months and 4 12 months after fracture). Costing The cost of fractures was estimated by multiplying the quantity of each different resource used by its corresponding value, i.e., unit cost. All costs are given at 2004 prices, inflated, when necessary, using the Swedish consumer price index [18]. The costs were converted from the Swedish krona (SEK) to euros ( ) using the yearly average exchange rate for 2004 ( SEK/ ) [19]. Medical care costs Cost of inpatient stays were assessed by multiplying the number of hospital bed days by the corresponding departmental daily cost per bed day. Unit costs for inpatient bed days, outpatient hospital visits, outpatient operations, X-ray examinations, primary care visits, physiotherapists and occupational therapists visits and telephone help were derived from the average unit cost for each resource, based on price lists for inter-regional care from 11 hospitals (Table 1). The costs for home visits (physicians, nurses and therapists ) were derived from a study by Janzon et al. [20]. Non-medical resource units and unit costs are shown in Table 2). Pharmaceutical costs The cost of pharmaceuticals was based on the cost of the average daily dose for each drug obtained from the

4 640 Table 1 Medical care resource units and unit costs (2004 prices, in euros) Resource unit Swedish national pharmacy [21]. Average daily costs and doses for the drugs that patients were prescribed are shown in Table 3. Community care costs Unit cost Reference source Inpatient care per bed day Orthopedics 700 [52] Geriatric 374 [53, 54] Emergency 427 [55 57] Other 545 [52] Outpatient care Primary care visits Physician 107 [53, 55 61] Nurse 52 [53, 58, 59, 62] Hospital outpatient visits Physician orthopedics 204 [53 56, 58, 59, 62, 63] Nurse orthopedics 76 [56, 58, 59, 62] Physician emergency 244 [53, 55 60, 62, 63] Nurse emergency 75 [63] Physician other 268 [54 56, 58, 59, 62, 63] Nurse other 65 [54, 58, 59, 62] Other visits to medical practitioners Occupational therapist 51 [56, 58, 62, 63] Physiotherapist 44 [56, 58, 62, 63] Outpatient operations Hip 625 [53, 56, 62] Wrist 646 [53, 56, 62] X-ray examinations Hip 60 [56] Vertebral 62 [56] Wrist 60 [56] Home visits Physician 283 [20] Nurse 71 [20] Occupational therapist 52 [20] Physiotherapist 52 [20] Telephone help Physician 44 [54, 56, 58] Nurse 22 [54, 56, 58] Costs for special living arrangements (nursing home, home for the elderly, group living) were obtained from the Stockholm municipality annual budget report 2003 [22]. Living at home was assumed not to be associated with any extra costs. Transportation costs for the disabled were based on the mean cost of transportation for the disabled in the Skåne region [23]. Table 2 Non-medical resource units and unit costs (2004 prices, in euros) Resource units Unit cost Reference source Living arrangements (day) Nursing home 176 [64] Home for the elderly 139 [64] Group living 160 [64] Home help per hour 25 [20] Transportation 16 [23] Work loss per day Men 178 [24] Women 152 [24] Informal care/hour (value of lost leisure time) 3 [27] Indirect costs and informal care The value of lost production, stratified by gender, was estimated from the average hourly pre-tax salary, including social insurance contributions in the private sector in 2003 [24]. The valuation of informal care is surrounded by uncertainty because no market prices for informal care exist. Therefore, different valuation principles can be applied Table 3 Average daily drug costs. Source: [21]. ATC The Anatomical, Therapeutic, Chemical classification system, COX2 cyclo-oxygenase-2, NSAID non-steroidal anti-inflammatory drug Average daily dose Average daily cost ( ) ATC group Bisphosphonates Didronate (etidronate) 400 mg 1.55 M05BA01 Fosamax (alendronate) 10 mg 1.14 M05BA04 Optinate (risedronate) 5 mg 1.19 M05BA07 SERMs Evista (raloxifene) 60 mg 1.01 G03XC01 Calcium and vitamin D supplements Calcitugg 1,000 mg 0.19 A12AA04 Kalcidon 1,000 mg 0.18 A12AA04 Kalcipos 1,000 mg 0.18 A12AA04 Kalcitena 1,000 mg 0.24 A12AA05 Ideos 1,000 mg 0.08 A12AX Calcium Sandoz 510 mg 0.15 A12AX Estrogens Trisekvens 1 tablet 0.24 G03FB05 Evorel 14 μg 0.22 G03CA03 Progynon 2 mg 0.13 G03CA03 Kliogest 1 tablet 0.12 G03FA01 Glucocorticoids 10 mg 0.11 H02AB Pain relievers a COX2 inhibitors 1.07 M01AH NSAIDs 0.30 M01A Opioids 0.40 N02A Light analgesics 0.19 N02B a Based on the average daily cost for the whole ATC group

5 641 when one is costing informal care. One way to estimate the opportunity cost of informal care is based on the income lost when relatives and others perform informal care [25]. The opportunity cost could then be estimated by the care givers wage rates. However, the time spent on informal care is often at the expense of both working time and leisure time. Another valuation principle is to value informal care at the market price of a close substitute, also known as the replacement cost method [26]. A close substitute for informal care could be, for example, in this study, home help. Because information concerning the employment status of informal care givers was not collected, informal care was conservatively valued at the cost of lost leisure time in the base estimations. Lost leisure time was valued at 3 per hour, which is the opportunity cost of travel time estimated in a contingent valuation study conducted by the Swedish Road Authority [27]. In a sensitivity analysis we also valued informal care using the replacement cost method, using the cost of home help to value informal care. Quality of life The quality of life was estimated through the EQ-5D questionnaire. This questionnaire is a general quality-of-life instrument that divides health status into five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression [28]. Each dimension is divided into three degrees of severity: no problem, some problems, major problems. The five health dimensions divide health status into 243 (3 5 ) possible health states. Social tariff values for these health states, estimated as time trade off (TTO) utility values, have been presented by Dolan et al. [29] This tariff was applied to the observed health states in the study. For the calculations of fracture-related loss of quality of life it was assumed that the patient would remain at the stated pre-fracture quality of life level if the fracture had not occurred. The annual loss of quality of life related to fracture was then calculated by subtraction of the prefracture quality of life from the estimated quality of life the year after fracture. The quality of life the year after fracture, the area under the curve, was estimated in two different ways. In the first way the quality of life was linearly interpolated between the point estimates at baseline, 4 months and 12 months. In the second way it was conservatively assumed that patients had reached their stated quality of life at 4 months as early as 1 month after fracture. The reason for this alternative calculation is that it might be reasonable to believe that the quality of life just after a fracture changes rather rapidly [17]. Information about the pre-fracture quality of life was collected from patients after the fracture and not before. This could potentially lead to some recollection bias, since the pre-fracture health status might have been perceived to have been better than it actually was. Therefore, the annual loss of quality of life was also estimated using Swedish age-differentiated EQ- 5D population norms [30] as a proxy for patients quality of life before fracture. Statistical analysis Cost data are often skewed and thus not normally distributed. The Shapiro Wilkes test was used for testing the normality of the cost and quality of life distributions [31]. If the data were not found to be normally distributed the nonparametric Kruskal Wallis one-way analysis of variance was used [31] to test the difference in costs and quality of life between different sub-groups (e.g., age, gender); if shown to be normally distributed the regular t-test was used [31]. If the data were not normally distributed, confidence Table 4 Patients characteristics and baseline information Characteristic Hip fracture (n=278) Vertebral fracture (n=81) Wrist fracture (n=276) Mean age (years) Age range (years) Proportion of women Proportion living in private residence before fracture Proportion having a previous: Hip fracture Vertebral fracture Wrist fracture Other osteoporotic fracture Multiple fractures Any osteoporotic fracture Proportion admitted the first day Proportion having first contact with emergency dept. Proportion having first contact with primary care Proportion having first contact with other Proportion working Proportion of workers that work full time Mean days from fracture to interview (standard deviation) Mean days from fracture to hospitalization if hospitalized (standard deviation) 4.24 (2.95) 0.41 (1.58) 6.56 (6.8) 0.19 (0.58) 8.08 (4.19) 0.96 (3.17)

6 642 intervals were obtained by the bias-corrected accelerated (bca) percentile bootstrapping method [32]. Because cost data are often heavily skewed they are not appropriate to use in regression modeling in their original shape. Therefore, in the analysis investigating the relationship between costs and other variables the Box Cox method [33] was used to find the best transformation of the cost data to fit a normal distribution. The transformed cost variable was then used as the dependent variable in a multivariate, ordinary, least-square (OLS) regression model. The Breusch Pagan test was used to test for heteroscedasticity [34]. In the presence of heteroscedasticity, White s corrected standard errors was applied [34]. All analyses were conducted with the statistical software package STATA 8.0 for windows, and 5% was used as the level of significance. Results Patients characteristics and baseline information Table 4 shows baseline information and patients characteristics for the 635 fracture patients (278 hip fractures, 81 vertebral fractures and 276 wrist fractures) included in the study. The proportion of women exceeded 78% in all three fracture groups, and the mean ages were 77.6 years, 75.9 years and 69.5 years for hip, vertebral and wrist fracture patients, respectively. Of the hip and vertebral fracture patients, 98% came from private residences before fracture, while all wrist fracture patients came from home. All the hip fracture patients were admitted to hospital when seeking care for the fracture, while 72% of the vertebral fracture patients were admitted and 9% of the wrist fracture patients. Approximately 20% of the hip and vertebral fracture patients and 9% of the wrist fracture patients had experienced a low-energy fracture during the 5-year period before being included in the study. The proportion of patients working was highest in the wrist fracture group. No patient above the age of 65 years had been working prior to the fracture. Resource use The utilization of medical care resources the year after fracture occurrence is presented in Table 5. Hip fracture patients have the greatest amount of hospitalization, with an average of 1.40 admissions per patient during the year after fracture, and wrist fracture the smallest amount, with 0.16 admissions per patient. The most frequent visit to a health professional for an average hip and vertebral fracture Table 5 Average utilization of medical care resources per patient the year after fracture Resource Hip fracture Vertebral fracture Wrist fracture Mean 95% CI Mean 95% CI Mean 95% CI Inpatient care Hospitalizations 1.40 ( ) 0.93 ( ) 0.16 ( ) Mean no. of days in hospital ( ) 7.96 ( ) 0.54 ( ) Outpatient care Primary care visits Physician 0.28 ( ) 0.51 ( ) 0.41 ( ) Nurse 0.16 ( ) 0.01 ( ) 0.27 ( ) Hospital visits Physician orthopaedics 1.18 ( ) 0.26 ( ) 2.01 ( ) Nurse orthopedics 0.01 (0 0.02) 0.01 ( ) 0.28 ( ) Physician emergency 0.72 ( ) 1.10 ( ) 1.01 ( ) Nurse emergency 0.00 (0 0) 0.00 (0 0) 0.01 ( ) Physician other 0.00 (0 0.01) 0.01 ( ) 0.02 (0 0.03) Nurse other 0.00 (0 0) 0.00 (0 0) 0.00 (0 0) Other visits Occupational therapist 0.17 ( ) 0.02 ( ) 1.68 ( ) Physiotherapist 3.98 ( ) 1.94 ( ) 1.69 ( ) Outpatient operations 0.00 (0 0.01) 0.00 (0 0) 0.22 ( ) X-ray examinations 2.29 ( ) 1.57 ( ) 2.73 ( ) Home visits Occupational therapist 0.85 ( ) 0.67 ( ) 0.04 (0 0.08) Physiotherapist 1.26 ( ) 0.77 ( ) 0.05 ( ) Nurse 0.99 ( ) 0.63 ( ) 0.05 ( ) Telephone help Physician 0.16 ( ) 0.35 ( ) 0.12 ( ) Nurse 0.12 ( ) 0.06 ( ) 0.05 ( ) Other 0.05 ( ) 0.02 ( ) 0.04 (0 0.07)

7 Table 6 Resource utilization per patient for community care, informal care and working status the last 4 weeks prior to interview at 4 months and 12 months after fracture Resource Hip fracture Vertebral fracture Wrist fracture Mean (95% CI) Mean (95% CI) Mean (95% CI) Four months Twelve months Four months Twelve months Four months Twelve months 643 Community and Proportion in special 0.05 informal care living accommodation ( ) Days in special living 1.41 accommodation ( ) (mean value) Hours home help per 1.87 week (mean value) ( ) Amount of transportation 1.26 last month (mean value) ( ) Hours of informal care 1.93 per week (mean value) ( ) Work related Proportion working 0.02 (0 0.04) Proportion working 1.00 full time (1 1) Sick days 0.46 (mean value) ( ) 0.05 ( ) 1.32 ( ) 1.42 ( ) 1.53 (1.07 2) 2.35 ( ) 0.01 (0 0.03) 1.00 (1 1) 0.07 ( ) 0.06 ( ) 1.57 ( ) 1.88 ( ) 2.40 ( ) 4.27 ( ) 0.06 ( ) 0.99 (0.97 1) 0.37 ( ) 0.06 (0.12 0) 1.63 ( ) 2.56 ( ) 2.47 ( ) 3.53 ( ) 0.06 ( ) 0.99 (0.97 1) 0.37 ( ) 0.00 (0 0) 0.00 (0 0) 0.21 ( ) 0.16 ( ) 0.34 ( ) 0.21 ( ) 0.96 ( ) 0.89 ( ) 0.00 (0 0) 0.00 (0 0) 0.07 (0 0.15) 0.24 ( ) 0.21 ( ) 0.21 ( ) 0.96 ( ) 0.48 ( ) patient was a visit to the physiotherapist, and, for an average wrist fracture patient, a visit to an orthopedic physician. In Table 6 the utilization of community care, informal care and working status is shown. Worthy of note is that vertebral fractures were associated with more transportation and a higher level of informal care than hip fractures, while wrist fracture patients utilized rather small amounts of municipality and informal care. Fracture-related pharmaceutical use at 4 months and 12 months is shown in Table 7. Vertebral fracture patients were, on average, prescribed most medicines compared to hip and wrist fracture patients both at 4 months and 12 months. Pain relievers were the most common medication, and, of the fracture risk-reducing treatments, calcium and vitamin D supplements, followed by bisphosphonates, were the most frequently used for all fracture types. Costs Average fracture-related costs the first year after a hip, vertebral and wrist fracture are shown in Table 8. Comparing the estimated cost items between the fracture types one can see that hip fractures were associated with the highest cost of inpatient care while wrist fractures were found to have the highest cost of outpatient care. Costs related to special living accommodation were approximately the same for hip and vertebral fractures, while wrist fractures were found to be associated with no costs for special living. Costs for other community care (i.e., transportation and home help) were highest for vertebral fractures. When the whole patient sample is considered, the estimated indirect cost was fairly low because few of the patients were below 65 years of age and working. Vertebral Table 7 Fracture-related pharmaceutical use at 4 months and 12 months after fracture Parameter Hip fracture Vertebral fracture Wrist fracture Four months Twelve months Four months Twelve months Four months Twelve months Proportion of patients on fracture-related medication Mean number of medicines per patient Proportion on calcium and vitamin D Proportion on estrogens Proportion on bisphosphonates Proportion on glucocorticoids Proportion on pain relievers

8 644 Table 8 Average costs the first year after fracture (2004 prices, in euros) Parameter Hip fracture Vertebral fracture Wrist fracture n=278; mean (95%CI) n=81; mean (95%CI) n=276; mean (95%CI) Medical care Inpatient care 8,805 (8,188 9,417) 5,533 (4,425 6,969) 357 ( ) Outpatient care 616 ( ) 527 ( ) 1,275 (1,187 1,385) Pharmaceuticals 170 ( ) 294 ( ) 124 ( ) Community care Special living accommodation 1,897 (1,049 3,154) 2,127 (906 4,744) Other community care 2,287 (1,832 2,948) 3,460 (2,397 5,142) 194 (98 395) Direct costs 13,775 (12,503 15,331) 11,941 (9,219 15,347) 1,950 (1,760 2,258) Indirect costs (loss of production) 93 (0 277) 156 (69 345) Informal care Valuation method Value of lost leisure time 353 ( ) 602 (316 1,297) 40 (23 65) Replacement costing method 2,824 (1,871 4,782) 4,815 (2,526 11,441) 321 ( ) Total cost a 14,221 (12,912 15,790) 12,544 (10,059 16,324) 2,147 (1,923 2,477) a Includes informal care costed using the value of lost leisure time fracture was associated with the highest community care and informal care costs, but, because of higher inpatientrelated costs, hip fracture had the highest total annual fracture-related cost ( 14,221), followed by vertebral fracture ( 12,544) and wrist fracture ( 2,147). Age- and gender-differentiated total costs for hip fracture are shown in Table 9. The estimated hip fracture costs vary significantly over the age groups (P<0.001) and range from 12,079 (50 64 years) to 18,214 (85 years and above). The hip fracture costs for women also vary significantly (P<0.001) across the age groups but not for men (P=0.35). The cost increase with increasing age among hip fracture patients is mostly because of a rise in resource use of community care, in particular home help and special living accommodation. Wrist fracture costs were not found to vary significantly over ages (P=0.88) or between genders (P=0.35). The cost of vertebral fractures did not differ between men and women (P=0.14). Below the age of 65 years the indirect costs were approximately 9% of the total costs for hip fractures and 23% for wrist fractures. Table 10 presents the age-differentiated fracture-related cost of vertebral fracture separated into patients that were hospitalized and those that were not when seeking care for the fracture. The cost was significantly higher for non-hospitalized patients 65 years or older than for younger patients (P<0.01). Hospitalized Table 9 Average age- and gender-differentiated annual cost of hip fractures (2004 prices, in euros) Group Number Direct costs Total cost Mean 95% CI Mean 95% CI years Men 10 13,369 (7,194 28,293) 14,824 (8,234 27,605) Women 13 10,077 (6,819 15,576) 11,082 (7,420 16,223) All patients 23 11,509 (8,177 17,831) 12,709 (9,242 18,428) years Men 18 12,194 (8,686 17,649) 12,399 (9,137 18,459) Women 53 11,746 (9,438 16,320) 11,868 (9,527 16,032) All patients 71 11,860 (9,981 15,416) 12,003 (9,936 15,324) years Men 22 16,517 (12,426 23,607) 17,051 (13,448 24,214) Women 96 12,487 (10,767 15,116) 12,830 (11,320 16,134) All patients ,238 (11,608 15,560) 13,617 (11,866 15,916) 85 years and above Men 9 11,436 (8,568 13,638) 14,859 (12,230 24,112) Women 57 18,558 (14,964 23,592) 18,743 (15,416 24,485) All patients 66 17,587 (14,564 22,403) 18,214 (15,199 23,445) All ages Men 59 13,889 (11,587 17,008) 14,920 (12,601 18,323) Women ,745 (12,311 15,728) 14,033 (12,549 16,095) All patients ,775 (12,504 15,331) 14,221 (12,912 15,790)

9 645 Table 10 Average age-differentiated annual cost of vertebral fractures (2004 prices, in euros) Age group All fractures Hospitalized Non-hospitalized Mean (95% CI) Mean (95% CI) Mean (95% CI) years 4,131 (1,507 13,144), n=13 9,015 (2,621 31,545), n=5 1,079 (567 2,861), n=8 65 years and above 14,152 (11,180 17,923), n=68 14,854 (11,793 19,894), n=53 11,675 (5,413 21,221), n=15 All ages 12,544 (10,059 16,324), n=81 14,350 (11,927 18,813), n=58 7,989 ( ,233), n=23 patients were found to have higher costs than non-hospitalized patients (P<0.01). There were no significant differences in any fracture cost between patients stated to have sustained a fracture 5 years prior to the study fracture and patients without a previous fracture. An econometric analysis was carried out to identify factors that can explain the difference in total costs the year after fracture for each fracture type. The independent variables that were considered for inclusion in the model were: quality of life before fracture, age at fracture, gender, and sustained fracture 5 years prior to study fracture. Regressions were estimated based on all patients and each fracture type separately. The Box Cox procedure suggested that the most appropriate transformation parameter value was close to 0, which is equivalent to a logarithmic transformation of the dependent variable. Therefore, a log-linear model was chosen for the estimations, i.e., the natural logarithm of the costs was regressed on the untransformed independent variables. The regression results are displayed in Table 11. In the regression estimation based on all patients the costs decreased significantly by approximately 7.6% for each 0.1 unit increase in quality of life before fracture and increased significantly (1.3% per year) for increasing age at fracture occurrence. Gender and prior fracture were not shown to have a significant impact on the total costs. Vertebral and wrist fracture-related costs were significantly lower (56% and 183%, respectively) than hip fracturerelated costs. Noticeable in the fracture-specific regressions is that quality of life before fracture and age at fracture occurrence were both significant in the hip fracture and the vertebral fracture regression, while no parameter was significant in the wrist fracture regression. Quality of life Table 12 shows the EQ-5D social tariff values before and after a fracture. EQ-5D was not available at all measurement points for one hip fracture patient and for four wrist fracture patients, and these patients were therefore excluded from the analysis. All three fracture types were associated with significant reductions in the measured quality of life after fracture compared to their perceived quality of life before fracture (P values <0.0001). The quality of life before fracture was highest for wrist fracture, which can partly be explained by the lower mean age of the wrist fracture patients. Vertebral and hip fractures are associated with the lowest quality of life at each measurement. After both vertebral and hip fracture, quality of life decreased to low levels directly after a fracture, but hip fracture patients showed higher quality of life levels at the 4-month and 12- month measurements than did vertebral fracture patients. Also shown in Table 12 are three different estimations (as described in the Methods and materials section) of the average quality of loss the year after a fracture. Comparing the three different estimation methods for each fracture type one finds that the simple interpolation method gives the highest reduction in quality of life for hip and wrist fractures, while the population-based utility approach gives the highest reduction for vertebral fractures. The assumption that patients reach their 4-month quality of life level at 1 month after fracture is the most conservative scenario giving the lowest reductions in quality of life for hip and Table 11 Linear regression analysis on (logarithmic) total costs Variable All patients Hip fractures Vertebral fractures Wrist fractures n=634 n=277 n=81 n=276 Coefficient P Coefficient P Coefficient P Coefficient P Quality of life before fracture < < Age at fracture occurrence < < Gender (0=men, 1=women) Prior fracture in the past 5 years Vertebral fracture <0.001 Wrist fracture <0.001 Constant < < < <0.001 R Robust standard errors Yes No Yes Yes

10 646 Table 12 Estimated health-related utility (EQ-5D social tariff values) (QoL quality of life) EQ-5D utility Hip fracture n=277 Vertebral fracture n=81 Wrist fracture n=276 Mean (95%CI) Mean (95%CI) Mean (95%CI) Perceived quality of life before fracture 0.80 ( ) 0.73 ( ) 0.89 ( ) After fracture 0.18 ( ) 0.18 ( ) 0.56 ( ) At 4 months 0.62 ( ) 0.47 ( ) 0.82 ( ) At 12 months 0.67 ( ) 0.49 ( ) 0.86 ( ) Average annual loss of quality of life Simple interpolation 0.23 ( ) 0.30 ( ) 0.10 ( ) Assuming the 4-month QoL reached after 1 month 0.17 ( ) 0.26 ( ) 0.06 ( ) Population-based utility values used as proxy for QoL before fracture 0.20 ( ) 0.34 ( ) 0.01 ( ) vertebral fractures. Wrist fracture was associated with the lowest loss of quality of life when the population-based approach was used. Multivariate ordinary least-square regression analyses were performed to analyze the impact of different factors that might influence the quality of loss after a fracture. The loss in quality of life, based on the assumption that patients reach their 4-month quality level at 1 month, is used as a dependent variable, and age, gender, hospitalization, quality of life before fracture, when seeking care for fracture, and sustained fracture 5 years prior to the study fracture were tested as independent variables. Regressions were estimated separately for each fracture type. The only variable that was shown to be significant (P values <0.01) for all fracture types was quality of life before fracture, which showed that the quality of loss increased with higher initial quality of life before fracture [parameter estimate 0.58 (hip fracture), 0.44 (vertebral fracture) and 0.75 (wrist fracture)]. In the wrist fracture regression a previous fracture was also significantly associated (P value =0.01) with a higher loss in quality of life. No other independent variable was shown to be significant. Discussion This study presents for the first time prospectively collected estimates on fracture-related costs and quality of life for the three most common osteoporotic fracture types. The design of the KOFOR study makes it possible to collect a broad range of costs from a societal perspective, e.g., costs for informal care. The cost of hip fractures have previously been assessed in Sweden [8, 10, 13, 14], but vertebral and wrist fractures have not. The pilot study preceding this study estimated hip, vertebral and wrist fracture-related costs, but the sample sizes were relatively small and did not include costs for community or informal care [17]. In two studies in 1998 and 1997, respectively, Zethraeus and Gerdtham and Zethraeus et al. [8, 10] estimated the annual extra cost of hip fracture. In the first study [10] the hip fracture costs for women ranged, over the ages, from approximately 8,400 to 22,600, and, for men, the hip fracture costs were from 6,700 to 20,400. In the second study [8], based on women in a private residence, the annual cost of a hip fracture varied, over the ages, from 11,700 to 38,000. Both studies were retrospective register studies estimating the extra cost of a hip fracture by comparing the costs the year after fracture with the costs the year before fracture. Neither study included primary care costs, indirect costs or informal care. Excluding these costs would give a hip fracture cost ranging from approximately 9,200 to 14,900 over the ages, based on our study sample, suggesting that the cost estimates for hip fracture in this study are somewhat lower than previously estimated. Reasons for this difference could be that the studies were undertaken at different times and because of productivity improvements in the healthcare sector. The number of bed days and days in special living accommodation has decreased from approximately 25 days to 14 days and from 28 days to 16 days, respectively [35]. The main explanation for this decrease in resource use is probably that the treatment of hip fractures has become more efficient since the beginning of the 1990s. The cost of fractures has also been estimated in some studies outside of Sweden [4, 9, 15, 16]. These cost estimates are hard to compare because both resource use in the treatment of fractures and price levels differ between countries. However, they confirm our results that hip fracture is the most costly fracture type, followed by vertebral fracture. Because of differences in resource use and price levels it is difficult to transfer cost estimates in this study to other countries. However, if one disregards differences in resource use and considers the price level only, it could be inferred that the Swedish estimated fracture costs in this study are probably quite high from an international perspective. For example, comparative price levels for healthcare [36] in Sweden are approximately 10% higher than in the UK, Germany and France, 20% higher than in Italy and 30% higher than in Spain. However, there are countries with healthcare price levels higher than Sweden s, such as Norway, Denmark, Switzerland and the USA. Previous Swedish estimates of the cost of vertebral and wrist fractures have been based mostly on expert opinion. Fracture cost estimates have varied from approximately 1,750 to 3,800 for vertebral fracture and approximately 440 to 2,080 for wrist fracture [37 39]. The cost of a vertebral fracture leading to hospitalization or no hospitalization was, in this study, calculated to be 14,330 and

11 647 7,989, respectively. The cost of a wrist fracture was estimated at 2,147. In particular, this suggests that the cost of vertebral fracture, previously, has been underestimated. Since a fracture is a strong risk indicator for another fracture, the proportion of patients receiving osteoporosis treatment for the fracture seems to be quite low. Only 32% to 54% were prescribed calcium and vitamin D, and 13% to 23% bisphosphonates, due to the fracture, 12 months after fracture. However, we do not know how many of the patients were receiving these drugs before the fracture, which makes it hard to assess how many of the patients were given osteoporosis treatment. An increased prescription rate for the patient group targeted in this study would probably lead to the prevention of more fractures at a relatively low increase in drug cost (e.g., the annual cost of bisphosphonate treatment is approximately 500). That is a cost that is likely, to some extent, to be offset by the prevention of more fractures among the patients. The impact of a hip fracture on quality of life estimated in this study confirms previous study results [17, 40 43]. In particular, the quality of life estimates are close to the values presented in Tidermark et al. [44] and Zethraeus et al. [17]. The quality of life values are close to the estimates found in Tidermark et al. [44], with the exception of the quality of life for hip fracture just after the event. The value in this study is lower than the one obtained in the previous study (0.14 vs 0.42). The reason for the difference may be that the quality of life just after a hip fracture changes rapidly, and that it then becomes important exactly when the measurement was carried out. This is confirmed by noting that the quality of life interview was carried out on average 4.24 days after the fracture, which is close to the fracture event. This suggests that the assumption that patients reach their estimated 4-month quality of life level after 1 month, when one is calculating the annual loss in quality of life, could be a reasonable assumption. The assumption of a rather rapid recuperation of quality of life after fracture is, however, not directly supported by our data when we compare the quality of life of patients interviewed within 4 days of the fracture and patients interviewed after 2 weeks. The quality of life for hip fracture patients was 0.13 for these groups, 0.19 (within 4 days) and 0.34 (after 2 weeks) for vertebral fracture patients, and 0.54 (within 4 days) and 0.55 (after 2 weeks) for wrist fracture patients. However, this analysis is somewhat uncertain because relatively few patients were interviewed 2 weeks after fracture. The estimated loss of quality of life related to vertebral fractures in this study is relatively large when compared with findings in other studies, such as those by Oleksik et al. [45] and Gabriel et al. [41] but is similar to estimates found in the pilot to this study [17]. Besides obvious differences in study design the main reason for this is that the quality of life assessments by Oleksik et al. [45] and Gabriel et al. [41] were based on patients that had sustained a vertebral fracture up to 5 years before answering the quality of life questionnaire, while our study enrols patients as soon as possible after the fracture occurrence. In clinical practice approximately one-tenth of the patients seeking care for a vertebral fracture is hospitalized [46]. This makes our study sample somewhat skewed, since 72% of the vertebral fracture patients were hospitalized. Patients that were hospitalized following a vertebral fracture had a lower quality of life before fracture than those not hospitalized. However, the annual loss in quality of life was similar for hospitalized and non-hospitalized patients. In a UK-based study by Dolan et al. [47] an annual loss in quality of life of approximately 0.02 related to wrist fracture was estimated from the EQ-5D questionnaire. This estimate is somewhat lower than our findings, at least when the patients stated quality of life values before fracture are used. The studies are not directly comparable, mainly because in the Dolan et al. [47] study patients were followed for about 40 days on average, while our study patients were followed for a year. A potential caveat with regard to the calculated loss of quality of life in this study is that the patients health status before fracture is retrospectively collected, i.e., the patients are asked about their perceived quality life before fracture after the fracture had occurred. This could probably lead to some potential recollection bias in the respect that patients might perceive their quality of life to be better than it actually was, which could lead to an overestimation of the loss in quality of life related to fracture. When quality of life norms for the Swedish population [30] were used as proxies for the quality of life before fracture, the annual loss was lower for hip and wrist fractures but slightly higher for vertebral fractures, implying that it is hard to say anything, at least when using this approach, about potential bias in the pre-fracture quality of life estimates. In our calculations of the annual loss of quality of life it was assumed that the quality of life estimated before the fracture would have been the same during the whole year if the fracture had not occurred. However, as can be seen from age-differentiated, health-related, quality of life estimates, the quality of life decreases with increasing age. This suggests that the loss in quality of life might have been overestimated in our calculations. From the study sample we can estimate a significant reduction in the quality of life before fracture of per 1-year increase in age. This would mean a negligible overestimation of of the quality of loss on an annual basis. There were no complete logs for patients that were considered for recruitment but not included. Thus, it is quite difficult to evaluate whether the patients included were worse off or better off than the average fracture patient. However, only patients that were judged to be able to participate fully in the study and did not suffer from any apparent co-morbid conditions were included in the KOFOR study; e.g., patients with dementia or other psychological problems were not included. If the aim is to estimate the average fracture cost for the average fracture patient in Sweden, the non-inclusion of some patients with high comorbidity might skew the cost estimates. It is reasonable to assume that the patients that were excluded were, to a large extent, already demanding substantial healthcare resources before fracture. The extra cost of a fracture would thus probably be fairly low, leading to an overestimation of the