Potentially inappropriate medication in the elderly in Germany: an economic appraisal of the PRISCUS list

The underlying anonymised aggregated prescription data were provided by the Scientific Institute (WIdO) of the German Local Health Care Fund (AOK). It was founded in 1976 and has the function to carry out research projects aimed at ensuring high quality and economic health care for the community of the Local Health Care Funds [15]. The WIdO has a prescription register of all drugs prescribed to patients in German SHI, excluding drugs dispensed at hospitals, over-the-counter (OTC) compounds, and prescriptions of private health insurance. Our study material is a register extraction of PRISCUS agents solely given to patients aged ≥65 years in 2009. The WIdO provided the sales volume, number of prescriptions, and the prescribed defined daily dose (DDD) for the top 30 PRISCUS agents with the highest sales volume. Additionally, the total sales volume, the total number of prescriptions, and the total amount of DDDs for all 83 agents on the PRISCUS list for the patient group ≥65 years of age were provided.

The data set was enhanced with additional information from the pharmaceutical prescription report from 2010 (PPR) [16]. The PPR is an annual publication by the WIdO about the 3000 most frequently prescribed drugs to the SHI population in Germany. Therefore, the provided data set and the enriched data have the same origin. First, we extracted the sales volume, prescriptions, DDDs, and costs per DDD of all PRISCUS agents for the entire SHI population from the PPR. Second, based on the enhanced data set, we calculated the share of patients ≥65 years of age in the entire SHI population for sales volume, prescriptions, and DDDs.

To calculate the potential costs of the recommended drug substitutions, we first derived the costs per DDD for the recommended surrogates from the PPR. In the first case, if a specifically drug was named, the costs per DDD were directly taken from the PPR. In the second case, if a drug group was recommended as an adequate alternative, a weighted average based on the costs per DDD and the prescribed annual DDDs was calculated. For example, the recommended alternatives on the PRISCUS list for indometacin are (1) paracetamol, (2) weak opioids (tramadol and codeine), (3) metamizol (after careful consideration of the risks and the benefits) and (4) "weaker" NSAIDs (e.g., ibuprofen) [9]. For (1) paracetamol and (3) metamizol, the costs per DDD were taken from the PPR (€0.49/DDD and €1.51/DDD). For (2) weak opioids, tramadol and codeine are explicitly recommended. Thus, the costs per DDD for these two agents were also taken from the PPR (€1.19/DDD and €2.65/DDD). By contrast, for (4) “weaker” NSAIDs, ibuprofen is only one example of a possible substitution. Thus, the costs per DDD for the whole group of "weaker" NSAIDs were calculated (€0.56/DDD).

In a second step, the reimbursement costs for an adjusted drug therapy was estimated under the consideration of the recommended surrogate according to the PRISCUS list. We assumed that the amount of DDDs for the recommended surrogates would not differ from the actual prescribed DDDs of PIM. To calculate the economic impact, the costs per DDD for the proposed substitution were multiplied by the prescribed amount of DDDs for patients ≥65 years of age. As several drugs or drug groups are recommended in some cases, a range of costs per DDD for surrogates had to be considered. Thus, we performed three different scenarios. Besides the cheapest (Scenario 1) and the most expensive surrogate (Scenario 2), we also calculated the mean costs per DDD of the proposed alternatives (Scenario 3) and the corresponding costs of the prescription.

Two systematic literature reviews were conducted to compare the results with national and international studies. The first one consisted of studies dealing with the prevalence of PIM in Germany measured by applying the PRISCUS list. The second one focused on international studies analysing reimbursement costs. We carried out two Medline searches on the 2^nd of November 2014. In order to identify the prevalence of PIM in Germany, we combined 18 synonyms for PIM, such as potentially inappropriate medication, inappropriate prescribing, and inappropriate practice in prescribing with “PRISCUS”. For the identification of studies of the costs of PIM, we combined the keywords cost OR costs OR cost analysis OR cost analyses OR drug reimbursement OR drug reimbursements OR economic burden OR economic evaluation and elder* OR old* OR geriatric* OR senior OR aged OR aging with the same words on PIM as in the search on prevalence. The complete search history for both searches can be obtained from the first author upon request. The search was not limited with regard to language and timeframe. Identified studies were searched by reviewing the references as well. For search 1, only studies of the prevalence of PIM in Germany were included. For search 2, papers were included if they examined the nationwide costs of PIM on an annual basis. Papers only examining the prevalence or the applicability of tools to identify PIM were excluded.

Our search yielded eight studies dealing with the prevalence of PIM in Germany, according to the PRISCUS list [12‐14, 17‐21]. The present study is based on prescription data in the ambulant stetting for all patients ≥65 years of age in German SHI, and results in a prescription prevalence of 4.8 % in 2009. The investigations of Amann et al. and Schubert et al. are most comparable regarding the database. Amann et al. calculated a standardised prevalence of 28.3 % based on data on 800,000 insurees [17]. A similar study design was chosen by Schubert et al. [13]. They used claims data from German health insurance providers to calculate a prevalence of 22.0 %. However, they both analysed the prevalence at the patient level not at the prescription level. Therefore, the comparison is limited.

The six remaining studies calculated prevalence based on smaller patient samples and in different study settings. The results lay between 16.6 % in an emergency department [14] and 43 % in geriatric rehabilitation [12]. For patients in nursing homes, a prevalence of 31.2 % was calculated [18], while for elderly people living at home the prevalence was lower at 18 % [19], 28.4 % [20] and 29.0 % [21]. Thus, the prevalence of PIM in the elderly in Germany is between one quarter and one third , but this depends on the study setting, possible limitations concerning indication, data collection, and the consideration of the limitations of the PRISCUS list (dose dependency and release form).

A search of economic analyses showed that limited information is available. In total, 12 publications dealing with the costs of PIM were identified, but none of them considered German SHI. The studies were conducted in the US [22‐26], Ireland [7, 27, 28], Japan [8], Finland [29], Saudi Arabia [30], and Switzerland [31]. They used different lists and mostly calculated the costs of PIM at the patient level [7, 8, 22, 25]. As our study addresses the annual economic relevance of PIM for SHI in Germany, only four studies from the US [24], Ireland [27, 28], and Finland [29] that addressed the national perspective are eligible for the comparison.

Bradley et al. analysed the PIM of patients >70 years of age according to the STOPP criteria in Northern Ireland, using data from the enhanced prescribing database of the National Health Service [27]. In total, costs of €6.1 million in 2009/2010 were calculated. This equals 5.38 % of total expenses on pharmaceuticals for people >70 years of age in Northern Ireland. Another study of Ireland by Cahir et al. published costs of €45.6 million for drug ingredients, tax, and pharmacy surcharges in 2007, representing 9 % of the overall expenditure on pharmaceuticals in people aged >70 years [28]. The calculation applying the STOPP criteria was based on the pharmacy claims of the National Shared Services Primary Care Reimbursement Service of the Health Service Executive. In the third study, Leikola et al. calculated reimbursement costs of €2.9 million for PIM medication to people >65 years of age in Finland in 2007, using the criteria of Fick and Beers [29]. This represents 0.7 % of total drug reimbursement for people aged >65 in Finland. The register-based cross-sectional study used data from Finland’s social insurance institution. The study from Fu et al., conducted in the US, calculated total health care expenditures of US$7.2 billion for community-dwelling patients exposed to PIM according to the criteria of Fick and Beers [24]. For medication, about 20 % of these costs were calculated. Consequently, costs of approximately US$1.44 billion were calculated for PIM in people >65 years of age in 2001 in the US.

In our study, the reimbursement costs of PIM were €387.8 million. Thus, the total reimbursement costs for PIM prescriptions in Germany are significantly higher than those in Finland, Ireland, and Northern Ireland, but lower than those in the US. The total sales of all proprietary mediclinical products in SHI were €28,499 million [16]. Thus, the proportion of PIM sales to elderly patients represents 1.36 %. However, a comparison of this number is hardly possible, as the studies from Finland, Ireland, and Northern Ireland used sales to elderly patients as the reference figure.

In general, the comparability of our results with other international studies is limited. First, different tools were used to detect PIM: Bradley et al. and Cahir et al. applied the STOPP criteria, which contain 65 criteria systematised according to the physiological systems to identify PIM [27, 28, 32]. The studies of Leikola et al. and Fu et al. used the criteria of Fick and Beer from 2003, which partly consider the diagnoses or conditions of patients [24, 29, 33]. In contrast to both tools, the PRISCUS list considers neither the physiological system nor diagnoses or conditions. Second, the lists were applied to different age groups: Bradley et al. calculated costs for pharmaceutical expenses for patients ≥70 years of age [27]. Our investigation covered prescriptions for patients ≥65 years of age. Third, different perspectives on costs were used in the studies: Cahir et al. summed the net ingredient costs, the value added tax, and the pharmacist dispensing fee [28]. In our study, costs per DDD are based on the sales data of each drug. Thus, the different approaches make a comparison of costs of PIM even more difficult.

Regarding PIM substitution, reimbursement costs are about to rise, independent of the scenario. If all prescriptions of PIM were substituted by the low-cost alternative, the total costs would be €325.9 million. Substitution by the most expensive alternative would lead to a total cost of €810.0 million. This gap can be explained by the enormous difference in the costs of the recommended therapeutic alternatives. For example, there are several different alternatives for etoricoxib (e.g., paracetamol, weak opioids (codeine, tramadol), metamizol, anticonvulsants (older anti-epileptics, newer anti-epileptics, neuromuscular blocking agents)). The cheapest alternative to etoricoxib (€1.21/DDD) is paracetamol (€0.49/DDD). The most expensive alternative is newer anti-epileptics at €4.03/DDD. The difference in costs per DDD might result in enormous cost differences in each scenario. By contrast, for solifenacin, which costs €1.43/DDD according to the PPR, the only alternative according to the PRISCUS list is trospium, with a cost of €1.05/DDD. Consequently, in both scenarios the substitution of solifenacin will decrease the total costs.

From a short-term economic perspective, there would be little incentive for SHI to support the substitution of PRISCUS agents. However, it has been shown that PIM leads to ADE which in turn can cause inpatient costs [14, 34, 35]. Furthermore, the application of PIM is directly associated with an increased risk of acute hospital admission [36‐41]. Moreover, in many clinical cases there may be no adequate substitute for a PIM. In these cases, the discontinuation of the PIM without providing a substitute will be the best choice from a clinical as well as from an economical perspective. Thus, from a long-term perspective, the additional costs of physician consultation and hospitalisation due to ADEs caused by PIM might decrease and compensate for the costs of substituting the PRISCUS agents.

Our research revealed several aspects to be considered in future research on PIM. The PRISCUS list was published in 2010. Our study is based on data from 2009. Thus, our study does not analyse the effect of the PRISCUS list and possible changes in the prescription behaviour of physicians in Germany, but serves as a reference for future comparisons. Further research should investigate if and to what extent prescription patterns and PIM costs have developed due to the PRISCUS list in order to show the influence and relevance for practical use.

Our investigation demonstrated that PIM is a relevant topic in drug therapy in the elderly in Germany. The elderly account for approximately 55 % of the sales of the 30 top-selling PRISCUS agents. Similar percentages were calculated for DDDs and prescriptions. The proportion of sales, DDDs, and prescriptions to the elderly is lower in all 83 agents of the PRISCUS list. The decrease underlines that only some of the 83 PRISCUS agents are really relevant. Thus, from an economic perspective and in terms of feasibility, a shorter PRISCUS list might be more applicable for the daily health service. However, the medical perspective has to be taken into consideration as well. A short-list should also include those agents with the worst risk/benefit ratios for the elderly.

There are limitations to this study as a result of the methods used. First, for several PIM the recommended substitutions may not be clinically adequate in a majority of clinical cases, while in other cases the recommended substitution would lead to off-label use. For example, herbal products, such as valerian, are often named as surrogates for psycholeptics (zopiclone, zolpidem and lormetazepam). From a medical perspective, however, it is questionable if a highly effective sedative can be completely substituted by herbal alternatives. Furthermore, valerian is a non-refundable OTC and no costs occur by substituting it. Consequently, in the low-cost scenario, approximately €42 million can be "saved" by applying valerian. Second, some PRISCUS agents are only potentially inappropriate, if a certain dose is exceeded or a certain release form is applied. This was not taken into account, as the data set of WIdO consisted of total prescriptions. Third, we only analysed the 30 top-selling drugs.

Another limitation is related to our applied scenarios. In general, the scenarios assumed that a PIM would be fully substituted by the cheapest, most expensive, or mean costs. However, a complete substitution might be unlikely in daily practice, which highlights the need for a calculation based on individual patient data. In general, the proposed substitutions have to be evaluated carefully by physicians. In some cases, groups of potential surrogates include other medication form the PRISCUS list. For example, the recommended substitution of amitriptyline are SSRI. However, the group of SSRI comprises PIM as well e.g. fluoxetine. It remains unclear if in those cases PIM might be substituted by another PIM as well. For our study, we included all recommended surrogates.

Not applicable.

Not applicable.

The datasets supporting the conclusions of this article are included within the article.