Our results suggest that prevalence of IMU is high among community-dwelling elderly, and that this use is associated with the female sex, advanced age and the number of drugs prescribed.
Studies clearly varied widely in their estimates of prevalence of inappropriate medication use by the elderly. Among estimates generated by applying several criteria, prevalence of IMU ranged from 11.5% to 62.5%. This variability may result from a number of factors, among them the diversity of inappropriateness criteria.
Most of the studies used classic explicit criteria, such as Beers. The Beers criteria, developed in 1991 using modified Delphi method, consists in a list of 30 drugs to be avoided in nursing home residents regardless of diagnoses, dose and frequency of medication use. Updates reflected the appearance of new drugs and knowledge, and broadened application of the criteria to ambulatory elderly [13
]. The latest version (Beers, 2002) considers 48 inappropriate medications or classes of medications regardless of diagnosis or conditions, and inappropriate medications or classes for 20 conditions. In 2001, Zhan et al. [19
] classified 1997 Beers Criteria drugs into 3 categories: "always avoid", "rarely appropriate", and "some indications". In 2003, an expert panel classified the 2003 Beers Criteria drugs into the same three categories, but only the categories "always avoid" and "rarely appropriate" were included in the HEDIS criteria [30
]. The McLeod method, which is also considered explicit, was developed by a Canadian panel of experts, and consists in 18 inappropriate medications for all elderly regardless of diagnoses or conditions, 16 inappropriate drug-disease interactions, and 4 inappropriate drug-drug interactions [20
]. Some studies in this review use more than one criterion [23
] or more than one version of the same criterion simultaneously [25
Most of the studies adapt the explicit criteria to exclude items that depend on dosage, use frequency, diagnosis, or the drug's availability in the country of the study. These adaptations are explained in part by the use of administrative databases containing no details about the drugs or how they are used. Also, extrapolations are made to countries other than where the criterion originated, where dosages may not be the same and prescription habits may be different from the method's country of origin. Also observed were adaptations to include drugs with a pharmacological profile similar to those mentioned in the criterion and available in the study country. These facts indicate the difficulties involved in extrapolating criteria from the country of origin to others. These difficulties are reflected even in the choice of study population, as was observed in this review: most of the studies analyzed (79%) were conducted in the USA, the country of origin of the Beers method, which also predominated in IMU analysis (74% of the studies). In addition to the difficulties regarding interchangeability of criteria, the explicit methods are criticized for their lack of specificity, given that they do not consider the characteristics or clinical condition of each patient [11
]. Accordingly, many authors prefer to include the term "potentially inappropriate" in their description of estimates.
In most of the studies that use multivariate analyses, IMU is associated with the female sex and advanced age (Table ). Also in the multivariate analyses, the number of drugs used or prescribed seems to be the most important factor associated with IMU. In Table the crude prevalence presented in single studies shows that polypharmacy is the covariate most strongly associated with IMU. This association suggests that the use of several drugs may also mean exposure to substances where the risks outweigh the benefits.
As regards the drugs/therapeutic classes most identified as inappropriate, the analysis of the medications was complicated by the heterogeneity of drug presentations. Classifications are not uniform and the rankings most used sometimes specify the drugs, sometimes the therapeutic classes, or even both drugs and therapeutic classes. Nonetheless, the studies do single out substances and therapeutic classes used for diseases highly prevalent among the world population, such as depression and anxiety [40
]. The medications used to treat these diseases in the elderly are present in several explicit methods and associated with severe adverse events, such as sedation, falls and cognitive dysfunctions [13
]. However, it is important to consider that only the short-acting benzodiazepines were strongly associated with fall-related injuries and that nowadays, the tricyclic antidepressants have been largely replaced by selective serotonin reuptake inhibitors because of lesser adverse effects [41
]. Prescription of medication to treat these diseases thus deserves close attention, given that withdrawal of such medication is associated with a reduction in adverse effects, and improvements in physical and cognitive functions in the elderly [42
This review was intended to contribute to knowledge about pharmacotherapy for the elderly by evaluating a non-institutionalized population. Our search strategy identified a large body of literature. Nonetheless, we may have missed relevant articles that were not identified, unpublished or excluded erroneously. Reliable evaluation of the vast and heterogeneous bibliographical material was assured by independent duplicate reading, and review by a third author at all stages of data selection and extraction. Certain limitations must be considered, however. Firstly, this review addressed only studies of administrative data sources, which are retrospective and have gaps in clinical information and in drug exposure data. On the other hand, they offer information on large populations. The number of articles published has been growing over the past few years. Contributions to the conceptual framework [43
] and statistical approaches [44
] have allowed a better understanding of the large administrative database as a valid means to examine the quality of medical services. Here, they were chosen for their representativeness, which yields more precise estimates and power to detect differences that otherwise would not offer statistical significance. As observed in Table the confidence intervals of estimates for the association between inappropriate medication use and sex, age or number of medications are very small. Studies of administrative data sources may also be useful as inexpensive screening tools in areas where quality can be investigated in greater depth. Lastly, this review did not address the repercussions of inappropriate medication use on the health of the elderly nor the capacity of the methods used to predict adverse outcomes, both of which are important considerations for clinical practice.
As more studies are published, it may be possible to measure and record all potentially important covariates. These should be considered in future studies in order to improve the ability to identify their impact on the estimates and develop control strategies. Variables such as sex, age and total number of medications used should be mandatory in future studies. It is also important to give attention to other sources of information, such as medical records and surveys, with a view to ascertaining to what extent different study designs entail discrepant results. Moreover, in the future, reviews of articles that analyze primary data from population surveys - with information on social variables, demographics, health status, diseases, lifestyle habits, and physical and mental limitations - can enrich our understanding of the complex network of factors involved in prescribing drugs for the elderly.