Introduction

Osteoporotic fractures are a major health problem in developed countries and have a corresponding impact on mortality, the quality of life and healthcare costs. Annually, 3.1 million osteoporotic fractures occur in Europe, equivalent to a loss of about two million disability-adjusted life years [1]. In Spain, it has been estimated that 714,000 osteoporotic fractures will occur in the next 10 years, mostly in women over 70 years old [2]. Targeting high-risk groups is a central component of cost-effective policies aiming at reducing the burden of osteoporotic fractures. Although different clinical approaches are currently available to lower the risk of fracture (e.g. lifestyle changes, the risk management of falls and several pharmacological treatments), there is considerable variability among clinical practice guidelines with regard to how to quantify that risk, who should receive densitometry testing, who should be treated, what treatment should be used and for how long [3–5].

The lack of strong, coherent, and widely accepted guidance for pharmacological treatment may result in a wide variation in the patterns of drug use among neighbouring populations with very similar demographic and epidemiological characteristics. While a significant amount of literature on Small Area Variations Analysis (SAVA) in surgical and medical hospitalizations and healthcare spending has been an essential element to interpret the behaviour of healthcare providers and to define public policies [6–9], studies examining geographic variations in pharmaceutical utilization have been scarce [10,11] until recent years [12–15]. Besides, very few have addressed–and then only partially–variability in osteoporosis medication utilization [16,17].

Beyond the evidence provided by clinical trials, the recommendations of clinical practice guidelines and the agreements reached by expert consensus panels, treatment decisions and the choice of a particular drug can be influenced by patient characteristics, physician and organizational factors, pharmaceutical promotion and healthcare system characteristics [18]. For instance, Spain is a country with one of the lowest incidence of osteoporotic fractures in the world, but utilization rates of osteoporosis medication are among the highest. In addition, considering that the treatment of osteoporosis is managed by primary care physicians and several other medical specialists (rheumatologists, gynaecologists and orthopaedic surgeons among others), to know to what extent the different levels of care explain geographical variations seems of special interest.

The aims of this study are: 1) to describe variations in utilization and expenditure on osteoporosis medication between the Primary Care Health Zones (PHZ) of the Valencia region, Spain, 2) to analyze observed variations using SAVA methods, and 3) to quantify the influence of the specialized care level on variations in utilization.

Materials and methods

Results

The average consumption of osteoporosis medication in the Valencia region in 2009 was 119.1 DID, ranging from 77.6 to 171.3 DID (a 2.2-fold difference) between PHZs in the 5th and 95th percentiles (Table 1). Biphosphonates represented 70.3% of the overall consumption (88.4 DID) followed by raloxifene (14.6%; 16.8 DID), strontium ranelate (8.6%; 10.4 DID), calcitonins (4.5%; 5.7 DID) and parathyroid hormones including teriparatide (2.0%; 2.4 DID). Regarding variability by therapeutic groups, biphosphonates showed the lowest variation (a 2.5 times higher consumption in the PHZ in the 95th percentile compared to that in the 5th percentile), with calcitonins and parathyroid hormones being the highest (around a 9-fold higher consumption between PHZs at the 95th and 5th percentiles). Age and copayment standardization (Table 1) did not substantially modify the variability among PHZs. Maps of ISR per PHZ show the geographical distribution of variation. In many cases the Primary Care Zones nested in the same Health Department adopt similar figures (S2 Appendix). Therapeutic group–specific Spearman correlations in standardized consumption (S3 Appendix) were significant and positive but weak (range: 0.14–0.35), except for two non-significant pairs (raloxifene/biphosphonates and strontium ranelate/parathyroid hormones).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Use of osteoporosis medication among women ≥ 50 years by Primary Healthcare Zones in the region of Valencia, 2009.

https://doi.org/10.1371/journal.pone.0199086.t001

The annual expenditure on osteoporosis medication (Table 2) was €59.7 million, equivalent to an expenditure of €66.4 per woman aged 50 and over. Expenditure varied 2.2-fold between PHZs in the 5th and 95th percentiles, ranging from €43.8 to €98.5. Biphosphonates, despite accounting for 70.3% of the overall consumption, only reached 53.1% of the overall expenditure on osteoporosis medication, while parathyroid hormones, representing only 2.3% of the overall consumption accounted for 18.3% of osteoporosis pharmaceutical expenditure. Biphosphonates showed the lowest variation in expenditure (2.5-fold between PHZs at the 95th and 5th percentiles), followed by raloxifene (EQ5-95: 3.6) and strontium ranelate (EQ5-95: 5.3), while calcitonines (EQ5-95: 8.6) and parathyroid hormones (EQ5-95: 9.9) showed the highest variation. Average prices of the DDD by therapeutic group (Table 3) showed very low or no variation, although they differed substantially between therapeutic groups, ranging from 1.1€/DDD for bisphosphonates to 14.3€/DDD for parathyroid hormones. The overall mean average price per DDD in osteoporosis medication was €1.5, showing low variation among PHZs (EQ5-95: 1.4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Osteoporosis medication expenditure in euros per woman aged 50 and over by Primary Healthcare Zones in the region of Valencia, 2009.

https://doi.org/10.1371/journal.pone.0199086.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Variance at the PHZ level explained by the Hospital Health Department in the region of Valencia, 2009.

https://doi.org/10.1371/journal.pone.0199086.t003

Regarding the standardized consumption in osteoporosis medication, HHDs explained a substantial part (39%) of the variance between PHZs (Table 3), ranging from 33% (biphosphonates or raloxifene) to 58% (parathyroid hormones) depending on the therapeutic group. However, the standard deviations between HHDs were lower than those within HHDs, except for parathyroid hormones where the within HHD standard deviation was higher than the between HHD deviation (0.47 and 0.40, respectively). Also, the high reliability values reflect a high intragroup correlation between PHZs from the same HHD. S4 Appendix shows the distribution of variation per HHD.

Regarding expenditure, the random-effects analysis of variance showed a similar pattern of results except for the overall pharmaceutical expenditure, where 54% of the variance between PHZs was explained by the HHD level. The standard deviation between (€12.9) and within (€11.8) HHD was nearly equivalent. The therapeutic group–specific standardized consumption explained 95% of the total expenditure on osteoporosis medication, with biphosphonates and parathyroid hormones being the most influential drugs (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Relation between ISR and total expenditure on osteoporosis medication.

Multivariate Regression Analysis.

https://doi.org/10.1371/journal.pone.0199086.t004

Discussion

Our study has several findings of interest. First of all, it shows considerable variability between PHZs, with ratios in the 95th percentile more than doubling those in the 5th percentile for osteoporosis medications altogether, and reaching up to a 10-fold difference for some costly drugs such as parathyroid hormones. These figures are higher than those reported in previous studies in Portugal (2004) [16] and Norway (2007) [17]. While Spain is one of the European and worldwide countries with a lower incidence of osteoporotic fractures [24,25], osteoporosis medications are among the most widely prescribed drugs (a recent report analyzing variability in the consumption of several drugs in 15 developed countries, including the U.S., Canada, and several European countries, placed Spain as the country with the highest utilization of osteoporosis medication [26]) and with temporal trends showing a very rapid growth in their utilization [27]. It is possible that greater utilization also implies greater variability.

Second, the magnitude of the differences in osteoporosis medication utilization between territories is barely modified by standardizing for age and copayment, suggesting a limited role for morbidity (age is a good proxy for the prevalence of osteoporosis) or cost-sharing as a source of variations between geographical territories.

Third, variation in drug utilization is transferred almost entirely to variation in expenditure per capita (in this case, per woman over 50) and furthermore the rates of utilization explained almost all the variance in drug spending. The different amounts of medications prescribed in each territory, rather than their price, are the main source of variation in osteoporosis medication expenditure. The explanation for this divergence should be sought in the homogeneity of prices in Spain of the most widely used osteoporosis medications (derived from the system of price regulation including the reference pricing system, and the existence of several drugs still under patent at the time of the study), which shifts the weight of the variability in expenditure to differences in utilization rates. This may not occur in other drug groups with a greater variety in prices (in countries with different systems of price regulation) or with increased consumption in high-cost drug groups. Finally, the public healthcare provision has an inflexible geographical organisation system in two levels of care with primary care health zones nested in hospital/specialized areas of service with very little across border flow. This becomes a particularly appropriate environment for analyzing the relationship between these two levels of care in drug pattern variations, and it is one of the aspects of interest in our study. The HHD level accounted for considerable share of the variation, with variations between HHDs being higher than variations within HHDs for all drugs except for parathyroid hormones. This is consistent with the fact that the prescription of this therapeutic class is mainly led by specialists (HHD level), in contrast to other classes such as biphosphonates, which is a drug class widely prescribed in both levels of care. A study conducted in the U.S. using Medicare drug and nondrug claims and bigger territorial demarcations (306 Hospital Referral regions [HRRs] and 3436 local hospital service areas [HSAs] for all the U.S.) found that 41% of the variation in drug spending between HSAs was explained by HRRs. A high proportion of low-spending HSAs were included in high-spending HRRs and vice-versa, limiting the potential of policy interventions at HRR level without affecting HSAs with differential patterns of behaviours regarding to the HRR in which they are included [15]. However, in the VHA where the managerial structures of primary and specialty care are integrated at the HHD level, these policies should have fewer limitations. In any case, the existence of PHZs with differential behaviour suggests that interventions aimed at reducing variations in care should consider the PHZ component, in order to avoid a potential restriction of health resources in populations that could suffer underuse problems, even if nested in a HHD with overuse problems. For instance, and on the basis of our findings with regard to the concerning variability in the use of parathyroid hormone and teriparatide, the Valencia Health Agency implemented in 2012 a system level intervention–both at HHD and PHZ levels—to rationalize the use of these therapies in the region [28].

Our findings are subject to several limitations. First, the ecological design limits both transfer of the results from the geographically aggregated population basis to the individual basis and the causal connections of the associations detected. Second, we used pharmacy claims data providing an accurate picture of dispensed drugs, which may be an imprecise measure of doctors’ prescriptions or of the medication truly taken by patients. Third, the Valencia region is a popular tourist destination and to some extent part of treatments–especially in coastal “sea and sand” PHZs–could have been filled by patients from other Spanish regions or indeed other countries. Fourth, indirect adjustment for age and copayment may be insufficient to control for differences in the risk of being treated with anti-osteoporotic drugs, and other variables not available in our study (e.g. people with malabsorption syndrome, treated with corticosteroids, with prior osteoporotic fracture, and other) could improve the adjustment. Fifth, we assigned all DDDs consumed to women aged 50 and over. Although this is a reasonable strategy because it is the population group that consumes the vast majority of osteoporosis medication variations between PCZs in utilization in men or young women could affect our estimates of variability for the group of women aged 50 or older, albeit in discreet manner. Last, the results presented correspond to 2009, and may not be generalizable to present-day Valencia (the use of anti-osteoporotic drugs in Spain may have been affected by several factors, from safety warnings about bisphosphonates to the expiration of some patents, warnings, restrictions of use and/or withdrawal of some drugs, the incorporation of new drugs, as well as policies developed in response to the economic crisis or the impoverishment of the population itself). A very recent study [29] shows trends in osteoporosis medication use until 2013, being the most striking a significant decrease in consumption during the economic crisis (just like it happened with drugs for other diseases). And, although the amounts consumed and its distribution among the different drugs may have changed, to our knowledge no other study to date has analysed the variability in osteoporosis medication utilization in the Spanish NHS nor in other settings, and thus, our study provides novel and contemporary information.

In recent decades a substantial body of evidence has highlighted the existence of wide geographical variation in health care utilization and spending that is not driven by patient characteristics and is not associated with the quality of care or patient outcomes [6–15]. This evidence has been interpreted both as a marker of inefficient resource use and as an opportunity for improving the quality and value of health care. Our study, focused on osteoporosis medication utilization and spending, shows results consistent with this previous literature. It has been suggested that clinical trials generated to obtain market authorizations produce a large body of evidence that, in turn, reduces uncertainty (and variability) in the use of drugs. But in many cases there is still significant unresolved uncertainty which results in variability in drug utilization in the real world. In the case of osteoporosis medication, previous studies suggest major disagreement among clinical practice guidelines and a massive use of anti-osteoporotic drugs in women under 65 years old with a low-risk of fracture, while there is significant underuse in women (and men) with a high-risk of fracture, including those who have already suffered a major osteoporotic fracture [30–33]. Large geographical variations in drug utilization rates could be useful to recognize inconsistencies in the clinical management of specific health problems and to guide policy proposals for targeting areas with problems of misuse. Our study suggests that these policies should be addressed to a system level component (including both the primary and the specialized level of care) and with specific strategies oriented to price and/or volume according to drugs and contexts.

Supporting information