Data Resource Profile: Danish online drug use statistics (MEDSTAT)

Denmark and the other Nordic countries were among the first to introduce publicly available drug use statistics.1 The first versions from the 1970s were mere aggregate statistics tabulating gross sales volumes per time unit for a number of drug classes. However, in 1994, the Danish Register of Medicinal Products Statistics (RMPS) was established based on detailed individual-level data on all prescriptions dispensed at Danish community pharmacies.2 Initially, the RMPS was only used for monitoring purposes; however, individual-level data became available for researchers in 2003.3 Based on the RMPS, the Danish Health Data Authority provides publicly available web-based drug use statistics.2 The MEDical STATistics (MEDSTAT) is the most commonly used of these web services,4 but recently two additional drug use statistics have been established, i.e. the Quarterly Drug Statistics5 (QDS) and the New Drug Statistics (NDS).6 All three sources provide measures based on individual-level patient data, resulting in an unprecedented refinement of public drug use statistics, and together they provide highly useful information for lay people, caregivers, legislators, administrators and researchers alike. Herein, we review the content, structure and research potential of the three drug use statistics, with main focus on MEDSTAT. The Danish National Health Service provides universal tax-supported health care, guaranteeing free and equal access to medical care by general practitioners and hospitals and partial reimbursement for prescribed medications.7 It is mandatory to report drug sales in Denmark, and consequently, the RMPS covers all national drug sales.2 Drug sales in the primary health care sector comprise purchases at pharmacies and, since October 2001, also purchases of over-the-counter (OTC) drugs at authorized non-pharmacy outlets. A sizeable proportion of OTC drugs, e.g. low-dose aspirin, are obtained by prescription due to co-prescribing with other drugs and partial reimbursement.8 Drugs prescribed by physicians at discharge from hospitals or at ambulatory contacts are dispensed from community pharmacies and thus registered as sales within the primary health care sector. A minor proportion of the total drug sales in the primary health care sector are derived from use by general practitioners, dentists and treatment centres in their practices.9 Drugs used in the secondary (hospital) health care sector comprise sales to hospitals or private clinics, which are registered in the RMPS at hospital and department levels.2,9 Pharmacies in Denmark are equipped with electronic accounting systems, which are primarily established to secure reimbursement at patient level from the National Health Service. Briefly, for each redeemed prescription, the patient’s unique civil personal register (CPR) number,10 prescriber identifier, product number, number of packages, retail price, reimbursement amount and date of drug dispensing are transferred electronically from the pharmacies to the RMPS.2 At the Danish Health Data Authority, additional information is added including the Anatomical Therapeutic Chemical (ATC) code defined by the World Health Organization (WHO),11 volume, package name, active substance, dosage form, strength, dispensing status and marketing holder. The volume for each package is specified by the number of defined daily doses (DDDs). The DDD is a WHO-defined statistical measure of drug consumption representing the assumed average maintenance dose required by an adult when the drug is used for its main indication. The DDDs for two closely related drugs should thus in principle be equipotent. The use of DDD allows for comparison of trends in drug utilization (nationally and internationally) independent of varying prices and pack sizes. For drugs with no WHO-defined DDD, the RMPS employs nationally defined DDDs, if available, or units such as ‘item’ (for vaccines) or ‘mg’ when a DDD is not applicable. In all available drug statistics, the latest updated DDDs and ATC codes are applied and updated retrospectively. MEDSTAT (www.medstat.dk)4 has provided aggregate statistics on medicine sales in the Danish primary health care sector since 1996 and the hospital sector since 1997.3,4 Complete individual-level data with information on age and sex first became available in 1999, because individual health insurance cards (with CPR numbers) for children aged below 16 years were first fully implemented in 1999, and drug use at nursing homes was not registered by the CPR number of the residents before 1999.10 Data in MEDSTAT are updated annually for the preceding calendar year, typically around May and latest in December. Thus, the backlog of annual statistics in MEDSTAT is 5–12 months. MEDSTAT has a unique presentation of all drug sales in Denmark at one website. Details are available for prescription sales, OTC sales (including sales from non-pharmacy outlets) and hospital sales. Information on specific drugs can be retrieved by three search modules: (i) Groups of medicines, (ii) ATC code and (iii) Product name. The Groups of medicines module provides data for sales within the primary health care sector and includes selected commonly used drugs (e.g. antidepressant and lipid-lowering drugs) and drug categories defined by combinations of ATC codes (e.g. oral contraceptives by generation and benzodiazepines) or dosage forms (e.g. hormone replacement therapy by systemic or local use). In the ATC code search module, drugs can be searched at substance level or in broader drug categories, defined by ATC codes. Searches by Product name include search options for product number, active substance and marketing authorization holder. Also, total sales of medicines are available at package level and the sales can be subdivided by health care sector and additionally, for the primary health care sector, on availability by OTC if any. MEDSTAT allows for stratification according to year, sex, age groups, geographical area (region) and sector (primary health care sector or hospital sector). The relatively broad default age groups can be changed to 1-year categories if needed. The ‘region’ variable refers to the Danish administrative regions, established in 2007, and includes the North, Central, Southern, Sealand, and Capital Regions of Denmark. The outcome variables include turnover, volume, number of individual users and proportion of total drug sale obtained by prescription. The denominator of the ‘prescription proportion’ is the total number of DDDs of the specific drug sold in the primary health care sector (including OTC sales), and the numerator is the number of DDDs obtained by prescription. Even for prescription-only drugs, the proportion obtained by prescription does not equal 100% because of the small bulk sale to private practices and medicine stocks at treatment centres. Danish online drug use statistics GP, general practitioner; no., number; OTC, over-the-counter. aAll based on the Register of Medicinal Products Statistics. bTypically prescription issued by general practitioners or outpatient specialty clinics. cAvaiblable in the MEDSTAT’s search module ‘ATC code’. dAvailable in the MEDSTAT’s search module ‘Product name’. eAvailable in the MEDSTAT’s search module ‘Groups of medicines’. fTurnover in the primary sector includes price, VAT (25%) and prescription charge (10 DKK per package). gWHO-defined daily dose (DDD), national DDD and other units (e.g. ‘item’ or ‘ml’). Danish online drug use statistics GP, general practitioner; no., number; OTC, over-the-counter. aAll based on the Register of Medicinal Products Statistics. bTypically prescription issued by general practitioners or outpatient specialty clinics. cAvaiblable in the MEDSTAT’s search module ‘ATC code’. dAvailable in the MEDSTAT’s search module ‘Product name’. eAvailable in the MEDSTAT’s search module ‘Groups of medicines’. fTurnover in the primary sector includes price, VAT (25%) and prescription charge (10 DKK per package). gWHO-defined daily dose (DDD), national DDD and other units (e.g. ‘item’ or ‘ml’). MEDSTAT’s interface for search modules and data output. Increase in use of sildenafil following patent expiration and price reduction in 2013 (figure modified from Quarterly Drug Statistics). The NDS closely monitors all purchases in the primary health care sector of drugs marketed within the past 10 years.6 Newly marketed drugs are defined as medicines with an ATC code new to the Danish Market, or a pre-existing ATC code with new active substance or new combination of active substances. At the NDS web page, individual drug sales can be retrieved by searches defined by ATC code/drug name or various outcome variables. Categorization by marketing year and reason for surveillance (new ATC code or combinations of active substances) are also feasible. Among monthly statistics provided by the NDS are number of unique new users, unique users stopping therapy, volume and turnover. As the reporting to the RMPS is mandatory, data in MEDSTAT are virtually complete, and the data accuracy is also high. Still, some caveats should be considered when interpreting data from MEDSTAT. Errors sometimes occur in the registration of prescription data, e.g. recording of wrong CPR numbers. Typically such errors are detected when data are cross-referenced by sex or age, e.g. contraceptives may be coded as purchased by young children in rare cases if the prescribing physician accidentally uses the health insurance card of a child (accompanying a mother) instead of the mother’s own. Table 2 Strength and weaknesses of MEDSTAT as a research data source Strength and weaknesses of MEDSTAT as a research data source The on-demand drug use statistics are particularly useful for researchers in the planning and dimensioning of new studies. Using information on drug prevalence in MEDSTAT, feasibility counts can be performed to evaluate whether a given study idea is worth pursuing. Such numbers can be incorporated into grant applications and study protocols to support the proposed study. MEDSTAT also allows for direct evaluation of statistical precision and sample size calculations. As an example, we conducted a case-control study on whether chronic use of valproate would potentially reduce cancer risk, since valproate had been discovered to inhibit histone deacetylase and had proven effective as an experimental antineoplastic drug for some cancers.14 Before embarking on the analysis, we tabulated the age- and sex-specific annual counts of valproate users based on data from MEDSTAT. As age- and sex-specific incidence rates of cancer were also available online,15 we were able to estimate the expected number of exposed outcomes under the null hypothesis (n = 75). This figure was considered the ‘bottleneck’ with respect to the statistical precision that we would be able to achieve, and we could thus estimate an expected coefficient of variation of our relative risk estimate of 12% (= 1/√75), assuming a Poisson distribution of the exposed cases.14 MEDSTAT also provides data for trends in drug utilization. In a recent study examining trends in utilization of non-steroidal anti-inflammatory drugs (NSAIDs) in Denmark during 1999–2012, we described the level of detail available from MEDSTAT on individual drug types/groups and stratification options.8 NSAIDs are ideal for illustration of the options in MEDSTAT because these agents are among the most commonly used drugs, comprise different types (notably aspirin vs non-aspirin NSAIDs), with different indications, include OTC agents, and have undergone major changes in usage during the coverage period of MEDSTAT. Of particular interest, we documented that the 1-year prevalence of prescribed low-dose aspirin increased throughout the coverage period, most notably among men.8 We also observed that non-aspirin NSAID use was frequent in all age groups above 15 years and had a female preponderance. The prevalence of prescribed non-aspirin NSAIDs decreased moderately after 2004, but substantial variation was seen between NSAID subtypes; the use of ibuprofen increased, use of all newer selective cyclooxygenase-2 inhibitors nearly ceased after 2004, diclofenac use decreased by nearly 50% after 2008 and naproxen use remained stable.8 Most recently, we have examined long-term nationwide trends in use of antithrombotic drugs in Denmark during 1999–2014. Using the number of DDDs per 1000 inhabitants as proxy for annual drug utilization, we were able to correlate trends in use of individual antithrombotic drugs with the time of publication of landmark studies advocating use of these agents.16 Patent expiration may effect drug utilization dramatically. As an example, the patent for Viagra® expired in the second quarter of 2013, with a subsequent instant reduction in average customer price and increase in total sales of sildenafil (Figure 2). The right to prescribe certain drugs has been expanded over time. As an example, the right to prescribe topical tacrolimus changed in 2008 from being restricted to dermatologists to also include general practitioners, and at the same time general reimbursement was introduced for tacrolimus. As shown from MEDSTAT, these changes increased the number of users almost 3-fold over the 5-year period from 2007 (n = 5230) to 2011 (n = 14 017). Studies of drug utilization trends may also specifically evaluate whether and to what extent drug use is influenced by issues or concerns raised by regulatory agencies or the media. Often, such issues relate to potential side effects of a drug, and a decline in drug use may reflect either that the physicians adjust their prescribing habits to the regulatory recommendations or that patients/parents decide to opt out of a treatment or vaccination for themselves/their children due to ongoing public debate. A particular phenomenon is a substantial decline in the use of a controversial therapy despite recommendations from physicians and/or regulatory agencies to initiate or continue therapy. As an example of changes in drug use caused by concerns about potential side effects, the decrease in diclofenac use after 20088 was likely a direct consequence of recommendations from the Danish Medicines Agency in 2008 and the Danish Society for Cardiology in 2009 to use diclofenac with caution due to an increased risk of cardiovascular disease.17 Consistent with also the most recent Danish and international guidelines for patients with or at high risk for cardiovascular disease,18,19 the annual number of diclofenac users per 1000 inhabitants is still declining (Figure 1). Prescribed use of the human papilloma virus vaccine outside the Danish Childhood Vaccination programme since 2011 (figure modified from the Quarterly Drug Statistics). MEDSTAT also provides information on the potential of prescription registries to capture individual-level drug use. As an example, NSAID use in Denmark is typically identified from prescription databases.24–26 However, it has been debated whether adequate information on NSAID use can be obtained from such databases.27 Stratification by27 or adjustment for28,29 aspirin use has been abandoned or dismissed in some studies arguing that, due to OTC availability, it is not feasible to reliably capture aspirin use in the Danish prescription registries.27 Using MEDSTAT, we were able to evaluate the veracity of this argument and we found that the vast majority of aspirin in low doses (mainly for cardiovascular prevention) is prescribed; the proportion of total low-dose aspirin sales prescribed on the individual level increased steadily from 62% in 1999 to 92% in 2012. We also examined prescribing proportions for non-aspirin NSAIDs. Overall, we concluded that OTC use of low-dose aspirin and non-aspirin NSAIDs is far less common in Denmark than in other countries,30 and thus that the potential for identification of NSAID use from prescription registries in Denmark is high. Because many studies in Denmark are conducted on registries with only regional coverage,24 regional differences could theoretically compromise the external validity of such studies. However, stratification on the most frequently used drug classes in MEDSTAT and calculation of coefficients of variation have revealed that there is no substantial heterogeneity between the five regions with regard to medication use or sociodemographic or other health-related characteristics.31 Permission to use MEDSTAT requires only reference to the Danish Health Data Authority, ‘MEDSTAT’ and date and the statement ‘Own calculations based on data from MEDSTAT’ in case of further calculations. MEDical STATistics (MEDSTAT) was established for drug monitoring and has provided free online nationwide statistics on drug use in the Danish health care system since 1996. Data are available by setting (primary or secondary health care sector) since 1997, age and sex since 1999 and geographical area (region) since 2007. Data are retrieved from the Danish Register of Medicinal Products Statistics and are accurate and virtually complete. Details are available for prescription sales, over-the-counter sales (including sales from non-pharmacy outlets) and hospital sales. Outcome variables include turnover, volume, number of individual users and proportion of total drug sale obtained by prescription. MEDSTAT is a unique resource for evaluation of study feasibility, sample size calculations, trends in drug utilization and the ability of the prescription registries to capture individual-level drug use. In queries requiring the most recent drug prescribing or close coverage of individuals initiating or discontinuing newly marketed drugs, MEDSTAT may be supplemented by two additional nationwide online drug use statistics (Quarterly Drug Statistics and New Drug Statistics). Permission to use MEDSTAT requires only reference to the Danish Health Data Authority. As this study was based solely on register data and did not involve any contact with patients, no approval was required from the Danish Scientific Ethical Committee. M.S. conceived the study idea and wrote the initial draft. All authors contributed to critical revision of the paper and agree to be accountable for all aspects of the work. Conflict of interest: None declared.