Title: Inappropriate Medication Use among the Elderly: a Systematic Review

s excluded? The authors state that the axes of the search equation were “population, intervention and outcome” yet none of the search terms pertain to intervention or outcomes. The authors do not present any results that pertain to the use of inappropriate prescribing criteria as an intervention or on the outcomes of using such drugs. It would appear that the authors have only conducted a review of the prevalence of inappropriate medication use in a highly selected group of studies based on secondary data sources. This is a major limitation and emphasizes the somewhat misleading nature of the title of this study. We did not exclude the rural population or frail elderly from the estimates of prevalence of inappropriate medications use among the elderly. We did however exclude articles that dealt exclusively with these population subgroups, so as to prevent biases and distorted estimates. The aim was thus to present estimates that might reflect, on average, what happens in the population over 60 years of age, considering a mix of different population strata. To make this clear, the second paragraph of the Methods/Selection section has been rewritten thus: “After reading the titles returned by the search, we excluded the following: case studies; studies in hospitals, nursing homes, or hospital emergency departments; studies of specific drugs or groups of drugs; studies exclusively of subgroups of ill, frail elderly or rural populations.” • Among the 10 titles selected without abstracts, 4 were news, 1 was practice guidelines, 1 was comment, 2 were a research project and were excluded for not offering data of interest to the study. • The Results section does not allude to “intervention and outcome” axes as mentioned in the Methods section, because the axes were solely a tool to make it easier to group the article search terms and not for studying interventions and outcomes on inappropriate use. For that reason, the interventions and outcomes on inappropriate use do not appear in the Results section. So as not to confuse readers, we excluded the second sentence of the first paragraph of the Methods/Search section “The search equation was formulated considering the axes population, intervention and outcome”. (ii) How did the authors assess the quality of the studies? Was there a minimum acceptable level of study design? Was heterogeneity explored? How were decisions made regarding suitability for meta-analysis? Were data pooled? Was sub-group analysis performed? • There are many consensus statements that have encouraged higher quality of reporting, including recommendations for randomized trials (CONSORT) (Deeks et al, 2003), observational studies (MOOSE) (Stroup et al, 2000) and observational epidemiological studies (STROBE) (von Elm et al, 2007). These are aimed at authors of reports, not at those seeking to assess validity (Sanderson et al, 2007). However, some points of the Strobe initiative (von Elm et al, 2007) have been considered here in order to evaluate the quality of the articles included in the review. Therefore, so as to make it clear what study quality criteria were evaluated, the information above has been included in the penultimate paragraph of the Methods section, in the following form: “Along the reading of the complete texts, data quality was also evaluated for inclusion in the review. Although the Strobe Initiative (von Elm et al, 2007) is not a tool for evaluating study quality, some points from it were considered here for that purpose, especially as regards the Methods section, as follows: (1) Setting: describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection; (2) Participants: Give the eligibility criteria, and the sources and methods of selection of participants; (3) Data search: give sources of data and details of methods of assessment (measurement).” -Deeks JJ, Dinnes J, D’Amico R et al. Evaluating non-randomised intervention studies. Health Technol Assess 2003;7:iii–173. -Stroup DF, Berlin JA, Morton SC et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA 2000;283:2008–12. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007; 335(7624):806-8 -Sanderson S, Tatt ID, Higgins JP. Tools for assessing quality and susceptibility to bias in observational studies in epidemiology: a systematic review and annotated bibliography. Int J Epidemiol. 2007 Jun;36(3):666-76 • In attempting to extract a measure that would summarize the prevalence estimates of IMU, we submitted the data to meta-analysis. Subgroups were thus formed according to criteria of appropriateness, sex and age using the Q-statistic to measure heterogeneity. Meta-analysis yielded a pooled weighted estimate of 25.5% (95%CI: 17.5-33.5), with significant heterogeneity (Q=1.3x106, df=17, p<0.001) which persisted after stratification by criteria of IMU, sex and age. The results discouraged the use of the method. (iii) The authors mention that “data were extracted from selected articles ... using a previously tested instrument”. What was this instrument? Is it valid and reliable? How was consensus achieved when discussing contentious issues with a third author – was this systematic according to pre-defined parameters? • The third paragraph of the Methods/Reading and data extraction section has been rewritten thus: “The data extraction form and the corresponding manual of instruction for completing it were tested initially with five articles and subsequently subjected to minor adjustments, such as including new data record fields or changes in format to accommodate the information recording better. It comprised seven sections which can be summarized as follows: identification of the article; description of the study source data base (type; country; scope); study population (individuals/visits/prescription); characterization of the participants (age; sex; schooling; income; co-morbidity); measures of frequency of inappropriate use (proportion of elderly); criteria of inappropriateness used (Beers, 1991; Beers, 1997; Beers, 2002; Drug Utilization Review; Zhan; McLeod; Medication Appropriateness Index, and others); medications (used; inappropriate by drugs/classes of drugs); associated factors (odds ratio; confidence intervals, p values). The form is available from the authors.” • Data were extracted from the selected articles by two authors independently. Doubts or disagreements between them relating to sampling, interpretation of estimates or statistics were resolved in meetings jointly with a third author who with more experience in pharmacoepidemiology. 5. Results (i) The authors state in the methods section that they excluded criteria that had been “adapted”. In the results section, the authors state that “12 of the 14 studies using Beers criteria adapted the criteria to restrict them to inappropriate drugs regardless of dosage or specific clinical conditions or even to drugs available in the country of study”. This is inconsistent and in the opinion of this reviewer, these constitute significant adaptations that would clearly impact on the measurement of the prevalence of inappropriate medication use. It is well established that studies which limit their measurement of inappropriate medication use to criteria “independent of diagnosis” underestimate the prevalence of inappropriate medication use by excluding highly prevalent and important drug-disease interactions in older people e.g. benzodiazepines with falls, calcium channel blockers with constipation, vasodilator medications with orthostatic hypotension etc. Exclusion by reason of adaptation of criteria mentioned in the Methods section (and which in the current version are in the Results section) has to do with studies whose criteria of inappropriateness differed widely from previously validated criteria and which could result in very discrepant estimates. As examples of far-reaching modifications, which justify exclusion, we mention: a) Anderson et al., 1997, which considered only 7 of the 30 inappropriate drugs described by Beers 1991; and b) Blackwell et al., 2009, which considers drugs described by Beers and others (not selected by consensus methods) that have the potential CNS adverse effects of dizziness/vertigo, drowsiness, and/or fainting. Anderson GM, Beers MH, Kerluke K. Auditing prescription practice using explicit criteria and computerized drug benefit claims data. J Eval Clin Pract. 1997 Nov;3(4):283-94. Blackwell SA, Montgomery MA, Waldo D, Baugh DK, Ciborowski GM, Gibson D. National study of medications associated with injury in elderly Medicare/Medicaid dual enrollees during 2003. J Am Pharm Assoc (2003). 2009 Nov-Dec;49(6):751-9. (ii) For the benefit of the reader, the authors should briefly allude to the key differences between Beers, Zhan and HEDIS criteria (mentioned in table 1). In the third paragraph of the Discussion/Inappropriateness criteria and factors associated section, we have included the text: “Most of the studies use classic explicit criteria, such as Beers. The Beers criterion, 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 contemplated the appearance of new drugs and knowledge and broadened application of the criterion to ambulatory elderly (Beers et al, 1991; Beers et al, 1997; Fick et al, 2003). 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 (2001), classified 1997 Beers Criteria drugs into 3 categories: “always avoid”, “rarely appropriate”, and “some indications”. In 2003, the 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 (Pugh et al, 2006). The method of McLeod, 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 (McLeod et al., 1997).” (iii) The results section does not allude to “intervention and outcome” axes as mentioned in the methods section. The Results section does not allude to “intervention and outcome” axes as mentioned in the Methods section, because the axes were solely a tool to make it easier to group the article search terms and not for studying interventions and outcomes on inappropriate use. For that reason, the interventions and outcomes on inappropriate use do not appear in the Results section. So as not to confuse readers, we excluded the second sentence of the first paragraph of the Methods/Search section “The search equation was formulated considering the axes population, intervention and outcome”. (iv) The authors describe the analysis of factors associated with inappropriate medication use i.e. female gender, age and number of drugs. This information is already published. The authors have not gone beyond this (i.e. meta-analysis). They have not truly analysed or commented on the strengths or weaknesses of these studies in the context or their research question. Although information on factors associated with inappropriate medication is already published, we fail to identify a review providing estimates from a set of articles (7) based on administrative data sources, like ours. Besides, this strategy allowed us to reveal great consistency among these studies with respect to the role of sex, age and multiple medication use as risk factors for inappropriate medication use. This seems to be the major contribution made by the article we submitted. 6. Discussion (i) The subgroup analysis presented in the discussion section should be presented in the results and the methodology should be described in the methods section. We have suppressed the paragraph that mentioned the meta-analyses, and the rationale is in Methods (ii), above. (ii) The authors should discuss the limitations of secondary databases as mentioned above. The secondary databases limitation is addressed in the first paragraph of the Discussion section, which was rewritten as: “Firstly, this review addressed only studies of administrative data sources, which reflect information on large populations. A growing number of articles have been published over the last several years. Contributions to the conceptual framework (Wray, 1995) and statistical approaches (Nebeker, 2007) have allowed a better understanding of the large administrative database as a valid means to examine quality of medical services. Here, the choice was made in view of the greater representativeness of secondary data, which result in more precise estimates and power to detect differences that otherwise would not present statistical significance. As we observed in Table 2, the confidence intervals of estimates for the association between inappropriate medication use and sex, age or number of medications are very small. In addition, studies of administrative data sources are costless and may be useful as screening tools in areas in which quality can be investigated in more depth. Otherwise, caution should be taken; studies with administrative databases as information sources are retrospective and have gaps in clinical information and in drug exposure data.” (iii) The authors conclude that “Educational measures contribute to improving pharmacotherapy”. This statement cannot be concluded from this review as interventions to improve pharmacotherapy were not evaluated. The authors should support such statements with references. The sentence was removed from the conclusions. Reviewer 4: Patrick J Barry