In this descriptive observational study, two collections of data were analyzed: all reimbursed medications prescribed to a sample of ambulatory elderly accessing two community pharmacies and, respectively, all medications recommended to a group of institutionalized elderly (urban facilities, Romania). All patients were over 65 years old at the time of the data collection and both data sets covered the same 30-day interval (March 2013).

A prescription could contain one to several medications indicated for various conditions.

The pharmacies were selected by convenience as they were considered to be representative for the majority of local community pharmacies in terms of accessibility to the elderly population and orientation towards patient's care (one had an independent- pharmacist owner and the other was included in a national chain of community pharmacies). Both offered services and medications to different population groups. For the ambulatory elderly, the analyzed data excluded details involving the use of nonprescription medications and natural products, as their use was not registered in the available data collection. For the institutionalized elderly, the analysis included all recommended medications (irrespective of their reimbursement status), following the indications of the primary-care physician who attended the facility. The medical files of the institutionalized elderly had variable levels of information related to medical history and laboratory investigations.

The analysis used three sets of PIM identification criteria: STOPP/START, the 2008 version (covering instances of misprescribing, overprescribing and underprescribing) and the PRISCUS list (misprescribing and overprescribing) [9, 10]. These sets of criteria are different in terms of need for patient's medical and medications history: the use of the STOPP/START tool frequently requires access to patients’ clinical information, while the PRISCUS list can be applied without such detail. The availability of the medications or drug classes mentioned in these evaluation criteria was considered as adequate for the descriptive purpose of the study: 63 (94 %) of the 65 medications or drug classes included in the STOPP criteria, 22 of the 22 (100 %) medications or drug classes mentioned in the START criteria, 45 of the 83 (54 %) medications or drug classes included in the PRISCUS list.

The study had a retrospective and observational character and formal consent was not required. The data analysis ensured patient’s confidentiality. The project received the approval of the local ethics committee. Two clinical pharmacists collected and analyzed the information available in the two data sets, using specific data collection forms. Descriptive statistics were used (Microsoft® Office Excel 2007, Microsoft Corporation).

For osteoarthritic pain, ambulatory elderly used various non-selective NSAIDs and coxibs, while the institutionalized elderly were exposed to diclofenac and ketoprofen administered intramuscularly. NSAIDs were co-prescribed with cardiovascular medications on 29 (85.30 % of 34) prescriptions and on 25 (80.64 % of 31) medical files presenting NSAIDs. 26 (76.5 % of 34) prescriptions and 27 (87.1 % of 31) medical files had NSAIDs recommended for more than 10 days/month, without concurrent therapy with a proton pump inhibitor. 24 (70.58 %) of NSAIDs prescriptions covered ≤ 20/30 days of analgesic therapy, in the absence of supplemental analgesic approaches. 21 (23.07 %) of the institutionalized elderly presented a diagnosis associated with a persistent pain syndrome, but lacked analgesic therapy. Acetaminophen was not recommended in the nursing home.

Typical and atypical antipsychotics were regularly used in the nursing home. 17 elderly with dementia (29.31 % of all 58 having a dementia diagnosis) received such therapy. Short to long-acting benzodiazepines were recommended for daily use in both environments of care. Vasodilators and circulation-promoting agents (pentoxifylline, Ginkgo biloba standardized extract, nicergoline, vinpocetine) were used by 26 (44.82 %) of the 58 institutionalized elderly with dementia and by 17 (26.98 %) of those 63 functionally dependent.

The items from the STOPP/START criteria requiring access to clinical information (medical history, medication history, degree of severity, intolerances or contra-indications etc.) were not applicable for the prescription analysis or for the majority of the included medical files, as the relevant clinical details were not available. For the ambulatory elderly, the STOPP/START criteria referring to the use of aspirin as an antiplatelet agent, of dipyridamole, loperamide or of calcium and vitamin D were not applied, as they refer to over-the-counter medications, not registered in the available database. Several medications from the PRISCUS list, although locally available, were not prescribed: ergotamine and its derivatives, muscle relaxants, baclofen, tetrazepam, clonidine, alpha-blockers, methyldopa, fluoxetine, ticlopidine, prasugrel, nitrofurantoin, sotalol, pethidine, phenobarbital.

This study identified several categories of PIM prescribed to two samples of elderly with different functional capacities and medical complexities, through the use of specific tools validated to this end: the STOPP/START criteria and the PRISCUS list. We found that 34.49 % of the prescriptions delivered to the ambulatory elderly and 82.41 % of the institutionalized patients were exposed to PIM. We decided to analyze these two elderly populations as they represent an appropriate environment for a more active pharmacist’s involvement in the complex process of geriatric care optimization. Although such pharmaceutical care activities are encouraged, their local necessity and benefit, in terms of safety and efficacy of elderly care, are to be demonstrated. Therefore, although our analysis covered a short period of time, it represents a first necessary step in a pilot project aimed at creating a screening instrument for PIM, specific to local prescribing practices and adapted for use by the local community pharmacists. Considering their potential clinical significance, our results could also initiate a pharmacist- prescriber dialogue aiming for a more geriatric- oriented process of care, in terms of both efficacy and safety. Furthermore, both the frequency and nature of the identified PIM suggest the need for further studies evaluating their likely impact on hospitalization rate or care- associated costs.

This is also the first study using the STOPP/START criteria and the PRISCUS list in conjunction, as explicit criteria for PIM identification in two samples of elderly patients cared for in two different environments. As suggested by previous similar research, the methodology used in this observational study tried to overcome the inherent limitations of explicit criteria, through the simultaneous use of three such European instruments [18]. Although several criteria referring to misprescribing instances, included in the STOPP tool overlapped with those of the PRISCUS list, their simultaneous use was considered as useful for the identification of the particular medication use patterns among the two subgroups. As the clinical information necessary for the application of these two sets of criteria is different (the STOPP criteria frequently refers to the patients’ medical or medication background, while the PRISCUS list does not), this aspect was considered appropriate, considering the fact that the prescriptions delivered to the ambulatory patients lacked the clinical detail, while this was present on the medical files of the institutionalized patients. Therefore, the limited access to the patient’s medical history or to information about the use of nonprescription medications proved to be an important although predictable obstacle in the use of all STOPP/START criteria referring to medications locally available. For example the STOPP criteria referring to the use of aspirin as an antiplatelet agent or the START criteria referring to the underuse of effective antihypertensive agents could not be used in the analysis of the ambulatory elderly therapy. This difference in applicability can partially justify the difference in terms of PIM frequency among the 2 sets of data. Moreover, this observation suggests caution in the interpretation of the results as the STOPP/START authors proved that misprescribed PIM are likely to be overestimated using STOPP and underprescribed PIM are likely to be underestimated using START, when the community pharmacist’s analysis lacked patients’ clinical details [19].

This higher exposure to PIM of the institutionalized elderly is a result confirmed in similar comparative studies that found institutionalization as a risk factor for PIP. Haasum et al. found that 30 % of the institutionalized and 12 % of the home-dwelling elderly were exposed to anticholinergic drugs, long-acting benzodiazepines, and concurrent use of 3 or more psychotropics [20]. Shah et al. confirmed this gap in PIM burden among the community and nursing home residents and also found a similar and significant difference in the number of daily medications administered to both populations: 4.9 in the community and 8.4 in care homes (3.22 and 8.26 daily medications, respectively, in our study), suggesting the increased risk for PIP as a consequence of the institutionalized elderly exposure to polypharmacy [21]. Polypharmacy is a prevalent geriatric phenomenon, correlated with increased care-related costs and greater mortality especially among those cognitively impaired [22]. The risk for polypharmacy is increased when medications with debatable benefit are being used, especially among those with limited life expectancy, for whom 40 to 50 % of the recommended medications can be considered as useless or overused [23]. From this point of view, the administration of vasodilators and various circulation-promoting agents (pentoxifylline, Ginkgo biloba standardised extract, nicergoline, vinpocetine) recommended as antidementia therapy to both populations, could be considered as potentially overprescribed. They represented 6.91 % of all ambulatory PIM and 27.14 % of all institutionalized elderly PIM, although the evidence of their benefit is reduced and concerns have been expressed regarding the potential for increased hemorrhagic risk, as stated by the PRISCUS criteria [10, 24].

The most frequent PIM categories identified were also different between the two environments of care: the underprescribed (55.34 %) and misprescribed (37.73 %) subtypes prevailed for the ambulatory elderly, while the misprescribed (62.14 %) and overprescribed (27.14 %) subtypes were more frequent in the institutionalized sample. Similarly to our study, Silva et al. used STOPP/START on a sample of institutionalized elderly and found a higher proportion of misprescribed PIM compared to the underprescribed cases: 76.82 % were STOPP criteria and 23.18 % were START criteria in their study [25]. The most prevalent examples were different nevertheless, as in our study the institutionalized elderly were mostly exposed to the high use of NSAIDs and to the underuse of antihypertensive therapy.

The misprescribed and underprescribed PIM identified through the use of STOPP/START criteria are comparable to others available in the literature. For example, a systematic review that included 12 prospective or retrospective observational studies and one randomized clinical trial, which applied full or modified STOPP/START on health records, found a variable frequency of PIM use, influenced by study design, ranging from 21.4 % to 79 %. Some of the identified PIM are similar to those of the present study: recommendations of long-acting benzodiazepines, benzodiazepines with long-acting metabolites, neuroleptics or underuse of statins in patients with documented history of coronary, cerebral or peripheral vascular disease [26]. Several studies that applied the PRISUS list of criteria were also identified. For example, one large-scale German analysis found that 25 % of the elderly received at least one PIM, with amitriptyline, acetyldigoxin, tetrazepam and oxazepam as the most frequent misprescribed PIM [27]. Reich et al. applied the PRISCUS list and the Beers’ 2012 criteria on health care claims data of four health insurers for managed care elderly in Switzerland, and found a 22.5 % estimated prevalence of PIM. PIM use was significantly associated with increasing number of chronic diseases or hospitalizations and again with polypharmacy [28].

The main PIM subtypes identified among the ambulatory elderly were the underprescribing of statins or β-blockers in coronary heart disease (47.16 % of all PIM), recommended to reduce mortality and morbidity even in the elderly population. The clinical relevance of these recommendations suggesting prophylactic approaches with a relative delayed benefit is however limited in the case of the frail elderly, frequently institutionalized, who could have different care needs [29]. Although the data available for the ambulatory elderly did not allow for the assessment of their degree of functionality, these results are similar to others showing a reduced use of prophylactic cardiovascular therapies or a trend of progressive reduction of use of cardiovascular therapies in nursing homes, especially for patients over 80–85 years old [21, 30]. Moreover, we found that 7.69 % of the institutionalized elderly could have benefited from the intensification of the antihypertensive regimen, as indicated by the START criteria, although this approach requires individualization in a population subgroup with a theoretically reduced life-expectancy [31].

The most frequent subtype of misprescribed PIM, identified in both environments of care, was the use of NSAIDs as analgesics in osteoarthritis, representing 56.66 % of total misprescribed -PIM in the ambulatory sample and 35.63 % of total misprescribed -PIM on the medical files, frequently associated with cardiovascular therapies or lacking gastro- protective agents. The amount of clinical information available for the ambulatory elderly could not allow for the assessment of the duration of use or for the presence of renal dysfunction in these patients, but the cardiovascular, gastrointestinal, central nervous system or renal risks remain a serious concern for potential safety- DRP [32]. The results are comparable to those obtained in similar studies that identified a reduced gastrointestinal protection among the elderly exposed to NSAIDs use [33]. The fact that for the institutionalized and generally frailer population, the potential misuse of NSAIDs represented the most frequent instance of PIP suggested the need for the reevaluation of the implemented pain management strategies. Furthermore, the NSAIDs use among other nursing home populations, described in the literature, was significantly lower (1.2 % or 3.8 %), as acetaminophen or opioids represented the preferred analgesic approach [21, 34]. The majority of products containing acetaminophen are available locally without a prescription and therefore they were not reimbursed, so their use could not be monitored in our ambulatory sample. Acetaminophen was not recommended to the institutionalized elderly so questions arise concerning its real use as an analgesic. Furthermore, it is difficult to ascertain the analgesic therapies used by the 71 % of the ambulatory elderly with persistent pain, having less than a 30 days duration recommendation for NSAIDs. Additionally, less than 8 % of the elderly from every sample used weak opioids and adjuvants. Although the underuse of pain medications is not a criterion in the START tool used in this analysis, our findings are in agreement with those identified through a European multicenter analysis, which identified a 48.4 % prevalence of analgesic underuse among institutionalized elderly [35]. Despite the fact that the recently published STOPP/START criteria versions 2 were not available at the time of our data collection, this recent version considers as a START criterion, the use of high-potency opioids in moderate-severe pain and underscores the need for efficient pain management in the elderly [36].

Long-acting benzodiazepines, zolpidem and zopiclone were identified on 4.64 % of the ambulatory elderly prescriptions and on 23.08 % of the institutionalized elderly medical files, for the management of both anxiety or insomnia. These molecules can have negative effects on the cognitive and motor functions, favoring falls and fractures with unproved benefits from long-term use [37, 38]. Bourgeois et al. and de Souto Barreto et al. identified a higher benzodiazepine use (50 % or 53.4 %) among the institutionalized Belgian and French elderly, correlated with the presence of pain syndromes and polypharmacy, but the administration of long half-life molecules or of unadjusted dosages represented potential DRP common to our subgroup [39, 40]. Among the ambulatory 70–89 years old Norwegian population, Neutel et al. found a higher benzodiazepine use that the one identified in our subgroup, with a 12.3 % prevalence of inappropriate use of benzodiazepines as hypnotics or anxiolytics [41]. Almost 20 % of the institutionalized elderly received antipsychotics, with haloperidol as the most frequently prescribed, while their use was infrequent in the ambulatory sample. Our findings are comparable to other similar investigations, as antipsychotic prescribing in Belgian nursing homes varied from 17.6 % to 32.9 % depending on study methodology [42, 43]. Antipsychotic use increases among the demented institutionalized elderly, having as potential indication the control of their neuropsychiatric symptoms or delirium episodes. Their use was associated with extrapyramidal, cardiovascular and cerebrovascular events and with an increased risk of mortality derived especially from the use of typical molecules [10, 44]. Both STOPP and PRISCUS criteria enlist the use of anticholinergic medications as unsafe in the elderly, but our data showed a limited prescription of these molecules (0.62 % of total PIM in the ambulatory elderly and 2.14 % of those institutionalized) by comparison with similar European or North American studies, that identified a 20.7 % and 73.62 % exposure in the institutionalized patients [45, 46]. These findings can be explained by a more reduced availability on the market of the anticholinergic medications presented in the criteria and also by different patterns of medication use in the elderly.

The limits of this study are a consequence of its retrospective nature and of the reduced scale of the data sets used for the analysis, which limit the generalization of these results. Medical history (including cognitive and functionality status) for the ambulatory patients was not available, nor was the information concerning the use of nonprescription medications or natural products. The available clinical detail concerning both collections of data was too limited to allow for the enquiry of several items included in the STOPP/START criteria, which require access to clinical information. The patient’s adherence to the recommended therapies was not known. For both environments of care, the analysis referred to the prescribed medications and not to those actually taken by the elderly patients.

Our study identified a potentially inappropriate use of NSAIDs as chronic analgesics in osteoarthritis both in ambulatory and institutionalized elderly, a reduced use of prophylactic cardiovascular therapies among the ambulatory elderly and a high overprescribing potential of vasodilators or circulation-promoting agents in the institutionalized sample. The necessity for a more consistent consideration of the risks implied by the aging process became evident. The results of our study encourage their confirmation through more complex investigations or more detailed analysis of the potential impact of the identified PIM on the rate of hospitalization or on care- related costs and could justify future interventional studies intended to optimize the elderly patient's pharmacotherapy.

DRP, drug-related problems; NSAIDs, nonsteroidal anti- inflammatory drugs; PIM, potentially inappropriate medications; PRISCUS, latin for “old and venerable”: START, Screening Tool to Alert doctors to Right, i.e. appropriate, indicated Treatment criteria; STOPP, Screening Tool of Older Persons’ prescriptions criteria.