A health economic analysis of osteoporotic fractures: who carries the burden?

A systematic literature search was performed, in accordance with PRISMA statement, to identify existing literature on COI regarding osteoporotic fractures. The search was conducted on September 25, 2012 in Cochrane Library, Embase, CINAHL, PubMed, EconLit, SveMed+ and Bibliotek.dk. Search terms, MeSH if possible, which were relevant to osteoporosis, fractures, costs and cost of illness, were applied, and equivalent Danish terms were used to search Bibliotek.dk. Inclusion criteria for this review were: men and/or women above 40 years; suffering from primary or secondary osteoporosis; hip, vertebral and/or wrist fractures; and cost of illness. The search identified 1,529 studies, and of these, 24 were identified as relevant. Six additional studies were identified from references, giving a total of 30 studies.

As osteoporosis is a chronic disease with recurrence of fractures, a Markov state transition model was appropriate to estimate the total societal cost of osteoporotic fractures in a Danish population of 50–99 years of age. Most orthopaedic surgeons often only diagnose fracture and not the cause of fracture (e.g. osteoporosis), thus to identify precise estimations of incidence of osteoporosis, fracture occurrence was combined with age [1]. From Danish literature, it was estimated that 17.7 and 40.8 % of Danish men and women over the age of 50 years are at risk of osteoporosis [1]. Thus, in 2011, the cohort should consist of 149,466 men and 388,474 women, which is the cohort used for Markov simulation in this study.

A Markov model was constructed and analysed with TreeAgePro version 2012, TreeAge Software, Inc. The model simulated the natural history of osteoporotic fractures by tracking fracture events and costs for the cohorts of patients as they transition across the ten health states, using 1-year cycles. The health states included in this model were: well, wrist, post-wrist, vertebral, post-vertebral, first hip, post-first hip, second hip, post-second hip and death. This is illustrated in Fig. 1. The post-fracture states were included to minimise the “memory-less” feature of Markov modelling, hence the increased risk of fractures following first fracture and associated cost were incorporated in the model simulating the natural history of osteoporotic fractures. All patients in the cohort were assigned to only one state for each cycle. The model was limited to maximum two hip fractures during a lifetime. One model for each gender was constructed to capture the different cost and probabilities across genders. Tables were applied for transition probabilities as these may change over time.

All transition probabilities are based on data from The Danish National Hospital Discharge Register (DNHDR) combined with relative risks for new fractures [12] and death [13] and were transformed to gender- and age-dependent probabilities.

The Markov model will be used to estimate incremental lifetime cost of fractures discounted back to 2011 for both genders and this will be divided into cost per health state and into payer perspective. The calculations for the total cost of osteoporosis will be based on incremental lifetime cost and the demographic composition.

The selected model population was all Danish citizens ≥50 years with an osteoporotic fracture between 2001 and 2010, as osteoporotic fractures most often occur in this age group [6, 14, 15]. The model population was divided into different age groups at 5-year intervals, ranging 50–99. An osteoporotic fracture was defined as wrist fracture (ICD-10: S52, S52.5B and S52.9), vertebral fractures (ICD-10: S22.0-S22.1, S32.0-S32.2 and S32.7-S32.8) or hip fractures (ICD-10: S72, S72.0 and S72.1), as these were considered directly associated with osteoporosis [2, 3]. For vertebral fractures however, the true prevalence is expected to be underestimated, since the majority are morphometric [16]. A Danish study has shown that only 10 % of vertebral fractures were treated in a hospital, 40 % received pain relief at own general practitioner (GP) and 50 % were not diagnosed [17]. Thus, the numbers identified by diagnosis codes only represent the minority of vertebral fractures, whereas it must be assumed that all hip fractures and wrist fractures were diagnosed and treated [18, 19]. In the Markov model, all vertebral fractures, both those with and without clinical manifestations, are included. Second fractures of the wrist and vertebral are not included in this model, but are considered as a new fractures, as patient specific data were not available. Furthermore, readmissions were not possible to exclude as patient-specific data were not available.

In Denmark, all citizens are covered by the national public health system and all contact to hospitals is free and registered according to their social security number. All contacts, both in- and outpatient, are registered in DNHDR, from which it is possible to locate the incidence of different diseases, e.g. fractures [19]. Studies investigating the validity of DNHDR found that the validity of fracture reports was 97 % [18]. From the DNHDR, data on diagnosis and operation codes were drawn for 2001–2010 to estimate the incidence of fractures, corresponding treatment and rehabilitation [20]. In Denmark, 25 % of wrist fractures were treated with surgery [17, 21], and 25 % were referred to physiotherapist [17, 22]. As many vertebral fractures are undiagnosed, probably only 10 % were registered in DNHDR [17], hence the remaining fractures and resource use were estimated based on literature and clinical experts. In case of hip fracture, 75.4 % of women and 82.3 % of men were surgically treated using internal fixation methods [4], and the remaining fractures were treated with alloplastic. Two kinds of rehabilitation were offered in Denmark with 96.3 % having standard rehabilitation [20]. On average, hip fracture patients were treated at rehabilitation institutes 2 h/week for 3 months [23].

Furthermore, the length of stay (LOS) for the different fractures was collected from DNHDR and used to estimate the productivity loss for the working age group (50–64 years) (LOS: wrist, 1 day; vertebral, 4 days) [20]. Productivity loss following hip fracture was set to 3 months, as patients were not able to resume their previous activities and work immediately after discharge from hospital [24]. The relative risk for new fractures differed between genders and was estimated based on the literature [12, 25].

In order to calculate age-specific incidence rates for fractures and death, the census database from Statistics Denmark was used to retrieve population numbers and deaths from 2001 to 2010 [26]. Data from Statistics Denmark were also used for data collection in order to calculate productivity costs (see Table 1). The risk of dying following a fracture was retrieved from literature [13, 27, 28].

Prescribed pharmaceuticals in Denmark are registered by the Danish Medicines Agency. The data cover 100 % of all prescriptions [29]. In Denmark, the GP must apply for reimbursement for each patients treated with anti-osteoporotic pharmaceuticals, except for alendronate in the case where a patient has had a low energy hip fracture [30]. However, these guidelines were only introduced recently and may thus have had little effect. It was presumed the persistence with the initial anti-osteoporotic pharmaceutical treatment was as stated in Roerholt et al. [31]. Calcium and vitamin D are neither prescribed nor reimbursed and purchases are thus not registered, so it was assumed that all patients using prescription pharmaceuticals are taking calcium and vitamin D supplement.

The Danish Longitudinal Study of Ageing consists of data from a national questionnaire survey performed in 1997, 2002 and 2007 on a sample of elderly Danish citizens aged ≥50 representative of their age groups. The database features a total of 23,704 completed questionnaires from 13,075 individuals. The three main areas of interest in this database are everyday life, labour market conditions and use of government benefits [32]. From this database, information regarding incremental use of personal care (16.6 %) and visits to GP (15.7 %) between osteoporotic and non-osteoporotic patients were retrieved. Hours per week of receiving personal care were estimated based on a Danish governmental report at 3.7 h [33].

Data regarding the use of assistive devices (walker, reaching aid, toilet raiser, dressing stick, crutch, bed raiser, bath bench and arthrodesis cushion) were retrieved from literature [34] and validated by experts. Further, 12 % of hip fracture patients received food service [34].

The following unpublished data from Odense University Hospital were kindly provided by M.D. Jesper Ryg: the incremental discharge to nursing home (i.e. compared to average Danish citizens) or discharge to respite care following hip fracture were 5 % for both. All future costs applied in the Markov model are discounted at 3 % annually as recommended by the National Institute for Health and Clinical Excellence [35].

Hospital costs are accumulated in secondary sector and do not differ between genders or age groups. Inpatient costs for the three fracture types were determined from Danish diagnosis-related group (DRG) browser [36], which contain information on total reported DRG cases with corresponding ICD-10 diagnoses, cost weights and average LOS [36]. For vertebral fractures, which are not diagnosed, no cost will be applied. This is illustrated in Table 1.

GP costs were accumulated in general practice and do not differ between gender or age groups. These included consultation fee, annual check-up, blood tests and telephone consultation [37].

Pharmaceuticals included prices for anti-osteoporotic pharmaceuticals, pain-relieving medication, calcium and vitamin D. From a calculated weight average purchase price for anti-osteoporotic pharmaceuticals, a patient would be reimbursed 75 % of purchase price [38]. This is illustrated in Table 1.

Rehabilitation, depending on type of fractures and the patient’s health state, during admission was included in the DRG. After admission, rehabilitation takes place at specialized units as standard rehabilitation or rehabilitation in own home. Respite care is temporary care, on average 26.8 days after hip fracture (unpublished data, M.D. Jesper Ryg), at a nursing home facility for patients needing care after hospitalisation. As many elderly eventually move to a nursing home, only the incremental length of stay due to osteoporotic hip fractures was included in this study. The incremental length of stay at nursing home was estimated at 2 years, as the average age for moving to nursing home in general population was 84 years [39] and for osteoporotic patients it was 82 years (unpublished data, M.D. Jesper Ryg). Patient co-payment for nursing home was fixed at 10–20 % of the total costs of stay at nursing home depending on the personal income. The patient’s co-payment to food at the nursing home and cost for staying at respite care was also included. This is illustrated in Table 1.

Social services and assistive devices were only included for hip fracture patients because it was estimated that wrist or vertebral fracture itself does not increase the need for social services or assistive devices compared to the general population. Cost data of social services was retrieved from the National Board of Social Affairs including costs of food delivery and personal care [40]. Personal care is a social service offered in Denmark and includes help for personal hygiene, getting dressed, etc. Eight types of assistive device were included in the study and costs for assistive devices were estimated from the municipal suppliers. This is illustrated in Table 1.

As a societal perspective was adopted, the cost of absence at work following osteoporotic fractures was included. The human capital method was used to estimate the productivity loss by valuing the time away from the labour market as the average daily wage in Denmark, according to gender and age groups [26]. To estimate the productivity losses, the number of days absent from work due to hospitalisation, rehabilitation, respite care and nursing home or premature death (before retirement age at 65 years) was included. Valuation of an average daily income and the labour force participation rate for men and women between 50 and 65 years were estimated from Statistics Denmark [26]. All costs are illustrated in Table 1. Cost of informal care was not included.

To assess the assumptions made for the COI model, several analyses were performed. Firstly, in order to confirm that the estimated number of fractures were the equivalent to those modelled by the programme, debugging was conducted. This debug method was done by applying tracker variables for the health states first hip, second hip, vertebral and wrist and comparing this result with the observed incidence from DNHDR. Secondly, to estimate which costs and probabilities would have the most impact on the COI, a one-way sensitivity analysis was conducted. For this analysis, intervals of likely best/worst scenarios were applied.