Introduction

According to the US Department of Health and Human Services, the proportion of the elderly in 2015 reached 8.5% of the total population and it is projected that it will continue to outpace that of a younger by reaching 1.6 billion globally over the next 35 years [1]. Due to multiple comorbidities with subsequent polypharmacy, and altered pharmacodinamic and pharmacokinetic drug properties, older people are at an increased risk of experiencing adverse drug events (ADEs) [2, 3], frequently as a consequence of potentially inappropriate medicines (PIMs) [4–7]. Most emergency hospitalisations for recognised adverse drug events in older adults resulted from commonly used medications, mostly anticoagulants, cardiovascular agents, antibiotics and oral hypoglycemic agents [8–10].

Studies have shown that PIMs may pose more risks than benefits to a patient, and that avoiding them renders a prudent way of reducing drug-related morbidity and mortality [11–15].

Since the introduction of the first criteria for inappropriate drug use by Dr. Mark Beers in 1991 [16], several screening tools have been devised and published in the USA, Canada and European countries [16–26]. Though one of the most frequently consulted sources about the safety of prescribing medications for older adults [16], Beer’s criteria clinical relevance and applicability was often adduced as ambiguous [27–29]. The lack of a reproducible, statistically significant association between Beers criteria PIMs and ADEs was highlighted in various trials [30], and its transferability was made questionable due to inclusion of many medications unavailable outside the USA. Therefore, various new criteria, both implicit and explicit, have been defined and validated for the purpose of minimising inappropriate prescribing in older people [31]. Increasingly used in Europe, STOPP (Screening Tool of Older People’s Prescriptions) criteria have recently been expanded and updated to include 80 items of more important PIMs significantly associated with ADEs [32]. Although STOPP criteria are extensively used in Europe and to some extent considered as ‘European Beers criteria’ [33], until recently, no assessment tool covered the drug markets of several European countries. To address this need and allow for the analysis of PIM patterns in and across the Europe, an expert-consensus list of PIMs covering the drug markets of seven European countries, called EU(7)-PIM list, was consented [34]. Moreover, by combining previously published and adopted tools, namely, adjusted Beers list [20], French consensus panel [22], McLeod’s list [18] and Lindblad’s list [21] of clinically important drug-disease interactions (DDIs), a new comprehensive tool that also includes the adjusted Hanlon’s [35] and Malone’s lists [36] of potentially serious drug-drug interactions was developed in 2012 [37] and subsequently tested [38, 39]. It was intended to be widely applicable to different healthcare settings and different geographical regions.

Availability and actual use of PIMs in the Belgian market as listed in the EU(7)-PIM list has recently been studied [40]. However, to the best of our knowledge, until now, EU(7)-PIM list has not been compared to any widely used criteria for detecting PIMs. Therefore, the primary aim of the present study was to measure the prevalence of PIMs by using the EU(7)-PIM list, STOPP version 2 criteria and new comprehensive protocol. Second, we wanted to screen the Croatian market on the availability of PIMs as listed in the EU(7)-PIM list, as to gain insights into remaining overlaps or gaps of PIMs available in Croatian national medication market, and therefore inform the evidence base with regards to the EU(7)-PIM list applicability.

Methods

Setting and participants

We prospectively studied 276 consecutive older patients who were acutely or electively admitted to the Department of Internal Medicine of the Dubrava University Hospital during the 15-month period from October 2014 to April 2016. The patients were admitted via the Emergency Department following referral by their general practitioner (GP) or self-referral or were transferred from another ward. The patients were eligible for the participation if they were aged 65 years or more with at least one regular prescription medication and signed informed consent. Patients’ data were retrieved by a careful review of patient’s admission documents, medical charts, hospital record and through the interview with the patient, or a patient’s family member or a caregiver, if needed. Supplementary information was obtained from the GP when necessary. Standard demographic details, current and previous medical diagnoses, regular prescription medicines, history of drug allergy and adverse drug reactions, serum biochemistry, renal function and blood pressure were collected for each patient by two independent clinical pharmacists. Non-prescription medicines were also included as the study focused on the prevalence of PIMs, and many over-the-counter drugs could have been potentially inappropriate (e.g. non-steroidal medicines) or of questionable efficacy (e.g. Ginkgo biloba). Anatomical Therapeutic Chemical (ATC) classification codes were used to analyse the prescribed therapy at hospital discharge. The principal diagnosis and comorbidities were coded according to the International Classification of Diseases (ICD-10 Version: 2016). The patients were excluded from the study if they were not able to answer the questions needed to complete the structured interview, did not have a caregiver or a proxy who could be interviewed if the patient was unable to participate in the interview, were unable or unwilling to give their consent. The study protocol was approved by the Hospital Ethics Committee. Presented data represents a secondary subset analysis of trial data evaluating the medication reconciliation process at hospital admission as a primary outcome measure (unpublished to date).

Potentially inappropriate medications

Identification of PIMs was based on patients’ medications at the point of discharge from the hospital, following modifications made by the admitting team, consulting physicians and clinical pharmacists involved in the study. Two clinical pharmacists collected the data simultaneously, thus no inter-rater agreement was needed. All disagreements or uncertainties were presented to the expert panel team which consisted of a clinical pharmacist and a clinical pharmacologist, which discussed the potential disagreements and reached a consensus regarding the identification of PIMs. To ensure adequate data collection, standardised data collection forms were used. The prevalence of PIMs prescribed to the elderly was detected using the following three screening protocols: (a) a EU(7)-PIM list [34], (b) the STOPP version 2 criteria (2014 version) [32] and (c) a new comprehensive protocol published in 2012 by Mimica Matanović and Vlahović Palčevski [37]. The EU(7)-PIM list consists of 282 chemical substances belonging to 34 therapeutic groups (i.e. 3-digit ATC codes; e.g. the nervous system) and recommendations for dose adjustments and therapeutic alternatives. EU(7)-PIM list PIMs are organised into two categories, independent of diagnosis (A) and considering diagnosis (B). PIMs in the updated STOPP version 2 criteria are organised by the physiological organ system class, and apart from explicit criteria, contain implicit prescribing rules that are drugs prescribed without a clinical indication, drugs prescribed beyond the recommended duration and duplicate drug class prescriptions. The new comprehensive PIM screening protocol consists of drugs with unfavourable benefit/risk ratio (33 criteria of individual drugs), drugs with questionable efficacy (6 individual drugs), drugs to be avoided with certain diseases/conditions (71 individual drug-disease interactions involving 28 diseases or conditions) and potentially serious drug-drug interactions (70 DDIs). In addition to listed PIMs and clinically important drug-drug interactions, alternative therapeutic solutions are suggested. It is believed by the authors that the tool consisting of PIMs and potential drug-drug interactions within the same protocol provides more comprehensive quality assessment of drug-prescribing behaviour to the elderly, which in turn may lead to better prescribing practises. As there are marked differences between the availability of drugs in different countries, authors of this protocol suggested that it may be advisable to use the protocol devised from the French or German screening tools together with Beers criteria, such as the comprehensive protocol. The variables were dichotomously coded, i.e. a patient had at least one PIM according to the EU(7)-PIM list, STOPP version 2 criteria or the new comprehensive protocol. The PIM index was calculated by dividing the number of identified PIMs by the total number of prescribed drugs, including over-the-counter drugs and herbal remedies, as it allows comparison of prescribing appropriateness between studies since the total amount of medications is also taken into account [41].

Furthermore, to check the availability, the official register of medications in Croatia (coded in ATC) was cross-referenced to the EU(7)-PIM list. Drugs unavailable in Croatia were excluded from the EU(7)-PIM list, STOPP version 2 criteria and the new comprehensive protocol.

Statistical analyses

The descriptive analysis of the study population characteristics included frequency distribution for the qualitative variables and measures of central tendency and dispersion for the quantitative variables. Independent t test was used to analyse the differences in variables between patients with PIM and those with no PIM. The number of detected PIMs per patient was compared between the three protocols using the repeated-measures analysis of variance (ANOVA). The Greeenhouse-Geissner adjustment was used when sphericity was not assumed and epsilon value was more than 0.8. Fisher’s LSD post hoc tests were used to determine the difference between each pair of protocols. The statistical evaluation of data was performed using IBM SPSS statistical software (v 20). Comparisons were conducted with a significance level at p < 0.05.

Results

Data were prospectively collected from 276 consecutive patients, of which 136 (49.3%) were female (Table 1). The median age (overall range) was 74 (65–92) years, with 44% of patients pertaining to a 75 years or older age group. Diseases of the circulatory system were the most prevalent pathology affecting 36.1% of patients, with hypertension as the most common presenting condition (10%), followed by diseases of the digestive system (11.8%) and endocrine, nutritional and metabolic diseases (10.8%).

Table 1 Study population characteristics
Full size table

The median number (overall range) of prescribed medications was 7 (1–17); 49.3% received from 8 to 17 drugs and polypharmacy (more than 4 prescribed medications) was recorded for 253 (91.7%) patients. Detailed information of the study population is provided in Table 1 and includes sociodemographic and clinical data.

According to the STOPP version 2 criteria, 393 PIMs were identified in 190 patients, thus representing 68.8% of the study sample. In total, 330 PIMs were identified in 184 patients (66.7% of all patients) using the EU(7)-PIM list, whilst 134 medicines were potentially inappropriate according to the comprehensive protocol and were distributed amongst 102 (37%) patients. The distribution of PIM indices per patient was 0.182 ± 0.181, 0.155 ± 0.151 and 0.062 ± 0,100 for STOPP version 2 criteria, EU(7)-PIM list and comprehensive protocol, respectively. Additionally, 252 potentially serious drug-drug interactions were identified according to the comprehensive protocol. These were distributed amongst more than half of the study sample (142 patients). Table 2 represents the percentage of patients receiving one or more PIMs or drugs involving serious interactions.

Table 2 Percentage of patients receiving one or more PIMs or drugs involving serious interactions
Full size table

The repeated-measures ANOVA showed that the number of PIMs per patient detected by either of the three protocols was significantly different; 1.20 ± 1.15, 1.42 ± 1.46 and 0.49 ± 0.76 PIMs per patient, for EU(7)-PIM list, STOPP version 2 criteria and comprehensive protocol, respectively (F = 93.1, p < 0.001). Fisher’s LSD post hoc tests showed that STOPP version 2 criteria identified significantly more PIMs than the other two protocols (STOPP version 2 criteria vs EU(7)-PIM list p < 0.001; STOPP version 2 criteria vs comprehensive protocol p < 0.001) and that the EU(7)-PIM list detected significantly more PIMs than the comprehensive protocol (EU(7)-PIM list vs comprehensive protocol p < 0.001).

According to all three sets of explicit criteria, benzodiazepines were the most frequently prescribed PIMs, followed by proton pump inhibitors and tramadol according to the EU(7)-PIM list and STOPP version 2 criteria only. Drugs most commonly resulting in inappropriate prescribing are shown in Table 3; listed drugs account for approximately half of all the identified PIMs according to all three sets of criteria. Additionally, according to the comprehensive protocol, the most frequently prescribed serious drug-drug interactions were those with the potential to result in hyperkalaemia, involving ACE inhibitors, potassium supplements or potassium-sparing diuretic (59 times); followed by combination of two antiplatelet agents or warfarin with antiplatelet agents (43 times) and two CNS agents (25 times) (Appendix A1). Moreover, STOPP version 2 criteria, which involve implicit prescribing rules, identified 23 drug class duplications and 22 drugs prescribed without an evidence-based clinical indication (Appendix A2), criteria not found in the other two protocols. The detailed PIM analysis according to all three protocols is shown in the appendices (Appendices A1, A2, A3, A4).

Table 3 Drugs most commonly resulting in inappropriate prescribing according to all three sets of criteria
Full size table

The number of prescribed drugs was the only variable associated with the number of identified PIMs for all three protocols; the mean number of drugs was significantly higher in the PIM-taking group of patients (p < 0.001) (Table 4). The patients taking PIMs were taking significantly more medicines (8.42 ± 3.248 vs 6.51 ± 2775 for EU(7)-list; 8.57 ± 3.217 vs 6.03 ± 2.466 for STOPP version 2 criteria and 8.84 ± 3.281 vs 7.16 ± 3.027 for the comprehensive protocol), confirming an increased risk for having a prescribed PIM in polymedicated patients. Moreover, gender, estimated glomerular filtration rate (eGFR) and number of comorbidities were associated with the proportion of PIMs, albeit according to the STOPP version 2 criteria only. Namely, PIMs identified according to the STOPP version 2 criteria were more common in women (p = 0.002), in patients with lower eGFR (p = 0.039) and more comorbidities (p = 0.001) (Table 4 and Table 5).

Table 4 Variables associated with potentially inappropriate medicines
Full size table
Table 5 Gender association with potentially inappropriate medicines
Full size table

Our additional aim was to check the Croatian market on the availability of PIMs as listed in the EU(7)-PIM list. Out of the 335 active substances listed in the EU(7)-PIM list, 125 (37.3%) were available in Croatia with 114 products being only available on prescription. The availability of PIMs from the EU(7)-PIM list on the Croatian market is shown in Table 6.

Table 6 The availability of PIMs from the EU(7)-PIM list on the Croatian market
Full size table

Discussion

This study was the first to compare the new EU(7)-PIM list with the other two validated and used criteria, namely, STOPP version 2 criteria and the new comprehensive protocol, and provide us with the insights regarding the prevalence of inappropriate prescribing to the elderly. Also, it was one of the first studies to have detected potentially inappropriate prescribing according to the EU(7)-PIM list [40].

The results indicate that potentially inappropriate prescribing is highly prevalent in hospitalised elderly patients presenting to the largest university hospital in the capital of Croatia, Zagreb, with 69% of patients receiving at least one PIM according to the STOPP version 2 criteria, and 67 and 37% receiving at least one PIM according to the EU(7)-PIM list and comprehensive protocol, respectively. This prevalence is comparable with the results of already-conducted research where the frequency of potentially inappropriate prescribing ranged between 44% for comprehensive protocol [38], 63% for STOPP version 2 criteria [42] and 73% for the EU(7)-PIM list [40]. Interestingly, these findings imply that the PIM prevalence detected by three different PIM screening tools coincides between different European countries, namely, Croatia, Belgium and Albania [38, 40, 42].

STOPP version 2 criteria detected significantly more PIMs in comparison with the EU(7)-PIM list and comprehensive protocol. A greater prevalence of PIMs identified by the STOPP version 2 criteria may be attributable to considerate amendments conducted in the updated version of STOPP criteria, thus rendering a higher sensitivity of these criteria. Namely, due to new emerging evidence, the total number of STOPP/START criteria increased by 31%. Hudra and coworkers compared the PIM prevalence identified by the two versions of STOPP criteria and found that the revised STOPP update identified significantly more PIMs (63.0% of patients) than the previous version (34.5% of patients) [42]. Furthermore, a rather high PIM prevalence identified by STOPP version 2 criteria in our study could be assignable to the three implicit prescribing rules, not found in the other two protocols, yielding a considerate proportion of identified PIMs. As opposed to explicit criteria, applied with little or no clinical judgement and developed from literature reviews, expert opinions and consensus techniques, implicit criteria focus on the patient, address entire medication regimen and primarily rely on expert professional judgement. Hence, more different types of drug therapy problems can be detected with the use of implicit criteria, as found in our study.

Encouraged by the work conducted by Wauters et al. [40], we cross-referenced the Croatian market on the availability of PIMs as listed in the EU(7)-PIM list, thus offering data not only on the actual use but also on the applicability of this tool in Croatia, a country not participating in constructing the abovementioned criteria. The overlap between the active substances in the EU(7)-PIM list and the official register of medications in Croatia was rather low, with 37% of medicines coinciding. Further research is warranted to repeat our exercise and confirm whether the EU(7)-PIM list is a suitable tool for detecting clinically relevant inappropriate prescribing. Even more, we strongly recommend future studies to focus on testing whether the EU(7)-PIM list is eligible for predicting adverse clinical outcomes, particularly rehospitalisation rates, side-effects, deterioration of clinical status etc.

Additionally, our study explored the association between PIMs and sociodemographic or clinical characteristics. The only variable associated with the number of PIMs for all three protocols was the number of drugs. As previously reported, PIM frequency is said to increase with the number of prescribed drugs [38, 42–47]. Stratifying by number of prescribed drugs, a significant upward trend in PIM prevalence was observed for each set of criteria in our study, implying that elderly polymedicated patients should be under special consideration. Moreover, only the STOPP version 2 criteria PIMs were significantly associated with gender, eGFR and comorbidities, suggesting it is a more sensitive tool. Namely, we believe that both eGFR and comorbidities are essential for appropriate prescribing in individual elderly patients and as such should be considered in PIM identification. In congruence with previous research [7, 39, 48], PIMs identified by the latest STOPP update were more prevalent amongst women, suggesting women were at higher risk of receiving a PIM.

The EU(7)-PIM list is restricted to drugs or drug classes, in some instances restricted to high doses (e.g. zopiclone used at doses higher than 3.75 mg/day) or prolonged treatment duration (e.g. proton-pump inhibitors used longer than 8 weeks) and can be used even if the clinical information available is minimal. Thus, it is suitable for pharmacoepidemiological application using administrative databases without any clinical information about the individuals concerned. It was argued by the authors of the EU(7)-PIM list that requiring clinical information for a comprehensive drug review of an individual patient is a disadvantage as opposed to applying a list which can be used with a minimum of clinical information. The researchers in this study strongly believe that a patient-specific approach is irreplaceable and that each patient should be viewed as a whole in a clinical context. Hence, the use of STOPP criteria renders a more comprehensive approach.

Additionally, a number of discrepancies between the three protocols were identified. Namely, a medicine was found as potentially inappropriate in one protocol and as a possible therapeutic solution in another (e.g. tramadol is mentioned as a PIM in STOPP version 2 criteria and EU(7)-PIM list as opposed to being a therapeutic solution for long-term use of full-dosage of non-steroidal anti-inflammatory drugs in the comprehensive protocol). Also, the main reason for pertaining to a PIM category can differ between the protocols (e.g. tramadol as a PIM due to risk of severe constipation in STOPP version 2 criteria or due to CNS side effects such as confusion, vertigo and nausea in EU(7)-PIM list). Furthermore, a medicine can be listed as a PIM in one set of criteria (e.g. the use of calcium channel blockers (of any kind) in patients with chronic constipation in EU(7)-PIM list and comprehensive protocol) and excluded from another as considered no longer completely accurate or relevant (STOPP criteria). However, regardless of the existing differences between the abovementioned protocols, all three of the protocols agreed on benzodiazepines as the most prevalent PIMs, thus repeating the results from previous studies [38–40]. Furthermore, we found that by intervening in six drug groups (Table 3), about 50% of potentially inappropriate prescribing could be avoided, according to any of the three criteria. We believe that this finding is very important because it emphasises the agreement between protocols on the most important PIMs in elderly patients and elucidates the areas of prescribing most attention should be paid at.

There were several limitations of this study. Our results are based on a hospitalised sample and therefore cannot be generalised to the entire elderly population. Such patients have multiple comorbidities and complex therapy regimens, hence are not comparable with community-dwelling older subjects. Second, the study sample was small and is not representative of the national population as a whole. Therefore, the results cannot be interpreted as reflecting the prescribing patterns and consequently the PIM patterns of Croatia.

In conclusion, the present study suggests that between one and two thirds of older hospitalised patients have at least one PIM at hospital discharge, depending on the tool employed for identification. STOPP version 2 criteria identified a significantly higher prevalence of potentially inappropriate medicines than the EU(7)-PIM list and comprehensive protocol and was found as a more sensitive tool for PIM detection. However, it is vital to evaluate whether the protocols are significantly associated with adverse clinical outcomes, as the evidence regarding PIMs’ impact on adverse clinical outcomes is scarce.