Review Article SUMMARY. Journal of Clinical Pharmacy and Therapeutics, 2013, 38, doi: /jcpt.12059

Transcription

1 Journal of Clinical Pharmacy and Therapeutics, 2013, 38, doi: /jcpt Review Article of the prevalence of potentially inappropriate prescribing in older adults, and evidence of clinical, humanistic and economic impact B. Hill-Taylor* BSP MLIS, I.Sketris* Pharm D MPA(HAS), J.Hayden PhD, S.Byrne PhD MPSI, D.O Sullivan MPharm MPSI and R. Christie BSc(Pharm) *College of Pharmacy, Dalhousie University, Halifax, NS, Department of Community Health and Epidemiology, Faculty of Medicine, Dalhousie University, Halifax, NS, Nova Scotia Cochrane Resource Centre, Dalhousie University, Halifax, NS, Canada, and Pharmaceutical Care Research Group, School of Pharmacy, University College Cork, Cork, Ireland Received 18 December 2012, Accepted 4 March 2013 Keywords: criteria, inappropriate prescribing, prescribing omissions, screening tool, screening tool of older person s potentially inappropriate prescriptions, screening tool to alert doctors to the right treatment SUMMARY What is known and Objective: Potentially inappropriate prescribing (PIP) has significant clinical, humanistic and economic impacts. Identifying PIP in older adults may reduce their burden of adverse drug events. Tools with explicit criteria are being developed to screen for PIP in this population. These tools vary in their ability to identify PIP in specific care settings and jurisdictions due to such factors as local prescribing practices and formularies. One promising set of screening tools are the STOPP (Screening Tool of Older Person s potentially inappropriate Prescriptions) and START (Screening Tool of Alert doctors to the Right Treatment) criteria. We conducted a systematic review of research studies that describe the application of the STOPP/START criteria and examined the evidence of the impact of STOPP/START on clinical, humanistic and economic outcomes in older adults. Methods: We performed a systematic review of studies from relevant biomedical databases and grey literature sources published from January 2007 to January We searched citation and reference lists and contacted content experts to identify additional studies. Two authors independently selected studies using a predefined protocol. We did not restrict selection to particular study designs; however, non-english studies were excluded during the selection process. Independent extraction of articles by two authors used predefined data fields. For randomized controlled trials and observational studies comparing STOPP/START to other explicit criteria, we assessed risk of bias using an adapted tool. Results and Discussion: We included 13 studies: a single randomized controlled trial and 12 observational studies. We performed a descriptive analysis as heterogeneity of study populations, interventions and study design precluded metaanalysis. All observational studies reported the prevalence of PIP; however, the application of the criteria was not consistent across Correspondence: B. Hill-Taylor, College of Pharmacy, Dalhousie University, 5968 College Street, P.O. Box 15000, Halifax, Nova Scotia, Canada B3H 4R2. Tel.: ; fax: ; murpp@eastlink.ca all studies. Seven of the observational studies compared STOPP/ START with other explicit criteria. The STOPP/START criteria were reported to be more sensitive than the more-frequentlycited Beers criteria in six studies, but less sensitive than a set of criteria developed in Australia. The STOPP criteria identified more medications associated with adverse drug events than the 2002 version of the Beers criteria. Patients with PIP, as identified by STOPP, had an 85% increased risk of adverse drug events in one study (OR = 185, 95% CI: ; P < 0001). There was limited evidence that the application of STOPP/START criteria optimized prescribing. Research involving the application of STOPP/START on the impact on the quality of life was not found. The direct costs of PIP were documented in three studies from Ireland, but more extensive analyses on the economic impact or studies from other jurisdictions were not found. What is new and Conclusion: The STOPP/START criteria have been used to review the medication profiles of communitydwelling, acute care and long-term care older in Europe, Asia and North America. Observational studies have reported the prevalence and predictors of PIP. The STOPP/START criteria appear to be more sensitive than the 2002 version of the Beers criteria. Limited evidence was found related to the clinical and economic impact of the STOPP/START criteria. WHAT IS KNOWN AND OBJECTIVE The health care of older adults presents significant challenges to, caregivers, healthcare providers and healthcare systems. Many factors contribute to this challenge, including (i) the increasing size of the demographic 1,2 ; (ii) the biological process of ageing, which may lead to increased sensitivity to drug effects due to changes in body composition (e.g. altered distribution volumes and altered permeability of the blood brain barrier), reduced ability to eliminate drugs (e.g. changes in liver metabolism and renal capacity), malnutrition and cachexia 3,4 ; (iii) the increased potential for and impact of comorbidities, multiple conditions and polypharmacy 5 ; and (iv) limited availability and access to appropriate evidence regarding drug effectiveness and safety in older and frail (i.e. studies have not been carried out, not synthesized or not in the public domain) John Wiley & Sons Ltd 360

2 Older adults exert a significant demand, due to disease burden on healthcare resources Potentially inappropriate prescribing (PIP) is reported to be highly prevalent in this age group and has been associated with adverse drug events (ADEs) leading to hospitalization and death. 12 In the United States, from 2007 to 2009, ADEs were responsible for approximately emergency hospitalizations of adults 65 years and older. 13 In Canada, the cost of ADE-related emergency department visits and subsequent hospitalizations for adults aged 66 and older was estimated to be $357 million per year in Inappropriate prescribing occurs when the risks associated with prescribing a medication outweigh the potential benefit of the medication in a particular patient. Various measures of PIP have been developed. For example, PIP may occur when medications are prescribed: (i) with no clear evidence-based indication; (ii) in higher doses or for a longer period than necessary; (iii) in combination with other drugs from the same drug class; (iv) in combination with other drugs that may lead to drug drug or drug illness interactions; (v) for who are susceptible to certain ADEs, for example, benzodiazepines in with a history of falls; and (vi) instead of a more cost-effective medication that is equally or more therapeutically effective. 15 Potentially inappropriate prescribing may also occur when a patient does not receive a medication indicated for the treatment or prevention of a disease or condition. 16 This may occur due to ageism, economic concerns, fear of adverse events or lack of prescribing knowledge. 4,17,18 Early detection of PIP may prevent ADEs and improve geriatric care. 6,19 PIP prevalence can often prove to be a useful indicator of prescribing quality. 7 In addition, evidence indicates that discontinuing inappropriate medications may improve subjective quality of life in older adults. 20 In 1991, Mark Beers and colleagues created a list of drugs that were considered to be inappropriate for use in older adults in longterm care facilities. 21 The Beers criteria have been revised and updated on a number of occasions since their development, with the most recent iteration being published in Although considered a cornerstone to the optimization of pharmaceutical care in older adults in the United States, 25 older versions have been criticized for their (i) limited transferability/applicability outside of the United States; (ii) failure to address a number of common PIPs (e.g. drug drug and drug disease interactions); (iii) failure to include criteria relating to potential underprescribing; and (iv) lack of user-friendly organization (e.g. by physiological systems). 15,26,27 As a result, at least 13 other explicit screening tools have been developed, many of which are specific to the country for which they were developed Many different clinicians including pharmacists, geriatricians, nurse practitioners, internists and researchers have used these different sets of explicit criteria to assess PIP across different healthcare settings and jurisdictions. 19,27 41 The 2012 iteration of the Beers criteria has addressed some, but not all of the criticisms listed above. 24 STOPP (Screening Tool of Older Person s potentially inappropriate Prescriptions) and START (Screening Tool to Alert doctors to the Right Treatment) are evidence-based sets of criteria, which were developed in Ireland using a modified Delphi process that involved 18 experts in geriatric pharmacotherapy from across the United Kingdom and Ireland. 42,43 STOPP consists of 65 criteria that help researchers and healthcare personnel systematically identify potentially inappropriate medications (PIMs). START, consisting of 22 criteria, identifies potential prescribing omissions (PPOs). The STOPP and START criteria may offer advantages over the Beers criteria and other screening tools. The criteria are organized according to the physiological systems to which each relate, thereby enhancing their useability. 42,43 In addition, rather than listing specific medications, which make transferability to different countries with different formularies more difficult, STOPP and START criteria refer to classes of medications. To date, the STOPP/ START criteria have been used by a variety of researchers across a number of healthcare settings in a number of different jurisdictions in Europe, Taiwan 47 and the United States. 48 There is a demonstrated research momentum towards the STOPP/START criteria, especially in the European Union. Our initial search on this subject found 17 records published in , 16 published in 2010, and 44 in In 2010, Levy and colleagues indicated that the STOPP/START criteria may be a good choice as an universal explicit criteria. 41 In 2012, a group of researchers and clinicians chose the STOPP/START criteria as the most appropriate choice for evaluating the prescribing of with multiple chronic conditions in Spain. 49 These tools have been proposed as the most appropriate for assessing PIP in the Netherlands. 50 The European Union Geriatric Medicine Society (EUGMS) have also recently announced their support for the STOPP/START criteria. 51 To our knowledge, no systematic review of the application and impact of STOPP/START has been undertaken. Although there is research establishing a relationship between PIP identified by STOPP/START criteria and other screening tools and adverse drug reactions, hospitalization rates and increased health services costs, this research is still in its infancy, and further study is needed to establish the full extent and true impact of this relationship. 37,52 This review informs researchers, clinicians and policy makers about the quality and extent of evidence relating to the STOPP/ START criteria. Specifically, it provides an overview of studies documenting STOPP- and START-identified PIP prevalence and outcomes in different healthcare settings and jurisdictions. In addition, with the development of electronic health records, administrative databases and computer-assisted order entry, integration of a reliable and validated set of criteria into these systems offers the potential for optimization of care with reasonable and sustainable expense and effort. Therefore, a systematic review of this nature is of particular value at this time. Our objective was to conduct a systematic review of research studies to describe the application of STOPP/START criteria and to examine the evidence of the impact of STOPP/START on clinical, humanistic and economic outcomes in older adults. METHODS Identification of studies Methods of the search strategy and inclusion criteria were specified in advance and documented in a protocol, which is available upon request. The methods of analyses were not predetermined due to the anticipated heterogeneity of the results. Randomized trials and non-randomized study designs investigating the impact and application of the STOPP/START criteria in adults aged 65 years and older were included. No language or publication restrictions were used in the original search; however, non-english studies and unpublished data were not included in the synthesis. Studies were eligible if published or accepted for publication between January 2007 and January 2012; studies only published as abstracts were excluded. The eligibility criteria 361

3 allowed access to all studies dated from the development of the criteria through to January Studies that used modified or a truncated list of STOPP/START criteria were considered for inclusion. Data from studies that involved the application of STOPP/ START to measure the prevalence of intervene in, or report predictors of, PIP were included. Indicators of the clinical and humanistic impact of the use of STOPP/START criteria on the patient and healthcare system (ADEs, physician visits, emergency department visits, hospitalization and quality of life) were developed by consensus and included in the protocol. Data demonstrating the economic impact of STOPP/START on PIP in older were also included. Biomedical databases and grey literature sources were systematically searched for published studies, prepublication presentations and abstracts. Keywords STOPP and START, STOPP Criteria and START Criteria with the limit Aged 65+ years (January 2007 to January 2012) were applied to Cochrane Library, Database of Abstracts of Reviews of Effectiveness (DARE), PubMed, EMBASE, CINAHL, ISI Web of Science, International Pharmaceutical Abstracts, Google Scholar, TRIP Database, ClinicalTrials.gov, metaregister of Controlled Trials (mrct), Pro- Quest Dissertation and Theses Database, and Index to Theses in Great Britain and Ireland. References from all included studies and systematic reviews of explicit criteria, which included STOPP/START, were searched. Citing articles, as identified by ISI Web of Science and Google Scholar, were examined. Experts in the field and authors of prepublication presentations, abstracts and registered clinical trials were contacted to identify studies that were in the process of publication. Authors of articles not available in English were contacted to see whether English translations were available. The search was conducted in December 2011 and was followed with RSS feeds from MED- LINE (PubMed) and EMBASE until January 2012 by one author (BHT). A title and abstract review of studies was performed independently in an unblinded standardized manner by two authors (DOS & BHT). Letters to the editor, commentaries and review articles were excluded. The full text of potentially relevant studies not eliminated by title and abstract was reviewed independently by both reviewers. Consensus was easily reached in the selection of included studies, and the two authors independently agreed to choose 13 of 77 citations. Data extraction A data extraction form was designed using Google Docs software (form available on request). The form was pilot-tested on two included studies and revised. Two authors independently performed the data extraction (DOS & BHT). One author checked extracted data for agreement (BHT). There were no disagreements in data extraction, beyond clarification of definitions and criteria. Information extracted included the following: (i) study design including methods and units of randomization, prospective vs. retrospective, characteristics of control group, data sources and time period; (ii) intervention details such as the professional designation of the health professional involved, criteria applied and a brief description; (iii) description of the location and setting, numbers and characteristics of participants including age, sex and numbers of medications prescribed; (iv) type of outcome measures reported; (v) results and key conclusions by the authors; and (vi) sources of funding. Risk of bias assessment Two authors (DOS, RC) independently assessed the risk of bias in eight of the included articles the randomized controlled trial (RCT) and observational studies that involved a comparison with another criteria. 17,46,52 57 The assessment was made using a modified quality assessment scale initially designed for studies of prognostic factors (QUIPS) 58 and was based on six domains: study participation, data collection, application of the STOPP/ START criteria, outcome measurement, study confounding, and statistical analysis and presentation. Each domain contained a checklist of three to nine components, which were used to render a score of low, moderate or high risk of bias for the entire domain. After independent review, both authors met to reconcile discrepancies in scoring each domain. Prior to discussion, reviewers initially agreed on 23 of 49 domain ratings (48%); 94% of domains were rated no more than one category apart with discrepancies commonly due to differing opinions on the potential severity of the possible bias. Consensus was easily reached in all assessments. We did not assess the risk of bias of cross-sectional studies that described the prevalence of PIPs using the STOPP/START criteria. Heterogeneity of study populations, interventions and study design precluded meta-analysis. Descriptive analysis was performed. PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) has been used in reporting this review. 59 RESULTS AND DISCUSSION Search results A search of relevant biomedical databases provided a total of 133 records: other sources including Google Scholar identified 31 records. Duplicates and records that did not meet the inclusion criteria were removed, leaving a total of 13 records: 12 observational studies 17,42,44,46,52,54,55,57 and a RCT. 55 (Fig. 1) Study characteristics Clinicians and researchers from the Republic of Ireland and Northern Ireland, including members of the original development team of the criteria, published eight of the included studies. 17,42,44,46,52,54,55,57 Gallagher was the lead author in both the only international comparative study examining PIP across six European jurisdictions 46 and the only RCT 55 that examined the effect that the application of the STOPP/START criteria could have on prescribing appropriateness and patient-related outcomes. Five other studies used the STOPP and/or the START criteria to evaluate PIP in older adults in Spain 53,56, the United States 48 and Taiwan. 47 Retrospective data were used in five of the observational studies. 44,47,48,53,60 This review includes the application of STOPP/START to the health records of approximately adults, all aged 65 years and older. (Table 1) The exact number of included is impossible to determine due to potential for crossover between the older included in the Cork-based hospital or long-term care studies and the database study. 17,42,44,46,52,54,55,57 The majority of participants were from the Northern Ireland and the Republic of Ireland (995%), reflecting both the large size of the study performed on the pharmacy database 44 and the smaller sizes of the other studies with non-irish participants ,53,56,60 The mean age of participants ranged from to 869 years 60 ; however, 362

4 Included Eligibility Screening Identification Records identified through database searching (n = 133) Records after duplicates removed (n =77) Records screened (n =77) Full-text articles assessed for eligibility (n =18) Studies included in qualitative synthesis (n =13) Studies included in risk of bias analysis (n =8) Additional records identified through other sources (n = 31) Records excluded (n = 59) Abstract only available (n = 19) Not published (n = 1) Editorials/commentary (n = 11) Reviews (n = 7) Not available in English (n = 6) Not specifically STOPP or START (n = 7) Validation studies (n = 5) Full-text articles excluded (n =5) Protocol only (n = 1) Age range of participants not 65+ years (n = 3) Only 10 criteria used (n = 1) Fig. 1. Flow diagram. some studies did not report mean age. 44,46,54,55 The majority of participants were female (from 53% 55 to 80% 56,60 ), except in two studies performed in veterans hospitals in Taiwan and the United States where 26% 47 and 1% 48 of the study population were female, respectively. The mean number of medications prescribed per participant ranged from 5 57 to in ten studies. 17,42,47,48,52,54 56,60 The majority of participants in the included studies were community dwelling (995%), who were assessed in either the primary care setting (community dwelling) 44,53,57 or while transitioning to a secondary care setting 42,46,48,52,54,55,57 ;05% of participants were divided between secondary care 47,53 and long-term care 17,53,56,60 (570 and 799, respectively). Five of the studies, including the RCT, applied the full STOPP and START criteria to participant s medication profiles, 46,47,55 57 three studies applied the STOPP criteria, 17,52,54 and one study applied the START criteria. 42 Four research teams used modified versions of the criteria. 44,48,53,60 Pyszka and colleagues noted that Medications without an appropriate diagnosis were the most common type of STOPP criteria that was identified, 48 although this is not one of the original STOPP criteria. 43 Conejos and colleagues 53 used the Spanish version of the criteria in their study. 61 Two of the studies applied a shortened/condensed version of the STOPP criteria to electronic databases without diagnoses or laboratory data being available. 44,60 The criteria selected for use in these studies were chosen on a consensus basis by an expert panel of five members in geriatric pharmacotherapy, clinical pharmacology, pharmacoepidemiology and academic general practice. 44 Cahir states that their research team applied 30 of the STOPP criteria. 44 Parsons states that their team used 31 STOPP criteria based on the list developed by Cahir and colleagues. 60 Prevalence of potentially inappropriate prescribing The observational studies reported the prevalence of PIP. 17,42,44,46 48,52 54,56,57,60 The prevalence of with at least one instance of PIP identified by the STOPP criteria ranged from 214% 57 to 79% 56 ; however, interpretation of this range should be made with caution due to the heterogeneity in both sample population and study design between the different studies. 363

5 Table 1. Characteristics of studies included in systematic review of STOPP/START criteria Source Study design Criteria applied Criteria applied by Compared criteria Population Country Data source # Age Sex Mean no. Rx/Patient Barry et al. Observational (2007) 42 prospective Gallagher & O Mahony (2008) 54 Observational prospective Ryan et al. Observational (2009) 57 cross-sectional Pyszka et al. Observational (2010) 48 cohort Conejos et al. Observational (2010) 53 prospective (hospital geriatric clinic) and cross-sectional (communitydwelling and longterm care) Cahir et al. Observational (2010) 44 cross-sectional Liu et al. Observational (2011) 47 cross-sectional Full START Physician None Communitydwelling being admitted to acute care Full STOPP Physician Beers (2002) 23 Communitydwelling being admitted to acute care Full STOPP and full START Modified STOPP Full START Modified STOPP Modified START Research pharmacist Clinical Pharmacist Independent observer Beers (2002) 23 Communitydwelling None Communitydwelling being admitted to acute care and discharged back to community care Beers (2002) 23 Communitydwelling, hospital geriatric clinic, and long-term care Partial STOPP Database None Communitydwelling Full STOPP and full START Gallagher et al. RCT Full STOPP and (2011) 55 full START Gallagher et al. Observational (2011) 46 prospective Byrne et al. Observational (2011) 17 cross-sectional Garcõa-Gollarte Observational et al. (2011) 56 cross-sectional Full STOPP and full START Physician None Acute care at discharge Physician Medication Appropriateness Index and Assessment of Underutilization index Consultants and senor residents in the field of geriatric medicine Full STOPP Research pharmacist Full STOPP and full START Physician with experience in the care of older persons Communitydwelling being admitted to acute care and discharged Beers (2002) 23 Communitydwelling admitted to multiple acute care facilities in 6 different countries Beers (2002) 23 Long-term-caredwelling in multiple facilities Australian Community- (Basger et al.) 28 dwelling, hospital or long-term care being admitted to multiple longterm care facilities Ireland Electronic database and paper-based medical records Ireland Medical records from community healthcare and acute care facility Ireland Electronic and United States paper-based medical records Electronic medical records Spain Electronic medical records and pharmacy claims database Ireland Electronic pharmacy claims database Taiwan Medical records from an acute care facility Ireland Patient or caregiver, community care and hospital medical records Switzerland, Ireland, Spain, Belgium, Italy, Czech Republic Patient or caregiver, community care and hospital medical records Ireland Electronic database and paper-based medical records Spain Electronic medical records 600 Mean (range): Median (range): 77 (65 94) 1329 Mean: (65 97) 111 Mean: (70+) 150 Mean: (69+) 56% Female Mean: 54 54% Female Mean (range): 62 (0 21) 61% Female Mean (range): 50 (1 19) 1% Female Mean: 129 (on Admission); 142 (discharge); 141 (Follow-up) 62% Female Not reported Not reported (70+) 57% Female Not reported 520 Mean: (65+) 382 Median (range), control 77 (65+); intervention 745 (65+) 900 Median (range): 82 (65+) 630 Mean (range): (65 99) 100 Mean (range): (65+) 26% Female % Female Control: 80, intervention: 74 61% Female Not reported 75% Female 78 80% Female

6 Table 1 (continued) Mean no. Rx/Patient Compared criteria Population Country Data source # Age Sex Criteria applied by Criteria applied Source Study design 600 Range: % Female 75 Ireland Patient or caregiver, community care and hospital medical records Full STOPP Not specified Beers (2002) 23 Communitydwelling, hospital or long-term care being admitted to acute care facility Hamilton et al. Observational (2011) 52 prospective 80% Female Mean (range): (69 106) Britain Medication Administration Records Partial STOPP Database None Long-term-care living with dementia in multiple facilities Parsons et al. Observational (2012) 60 cross-sectional STOPP, screening tool of older person s prescriptions; START, screening tool to alert doctors to right treatment. Rx prescriptions. The observational studies also outlined the most commonly encountered instances of prescribing potentially inappropriate medications 17,42,44,46 48,52 54,56,57,60 : (i) PPI for peptic ulcer disease at full therapeutic dosage for > 8 weeks 17,44,46,52,56,57,60 ; (ii) longterm (i.e. >1 month), long-acting benzodiazepines and benzodiazepines with long-acting metabolites 17,44,46,47,52 54,56,57 ; and (iii) longterm (i.e. >1 month) neuroleptics as long-term hypnotics. 46,47,56,60 The START criteria identified at least one instance of PPO in 227% 57 to 74% 56 of. Seven studies outlined frequently encountered PPO instances 42,46 48,53,56,57 : (i) calcium and vitamin D supplement in with known osteoporosis 46,48,56,57 ; (ii) statin therapy in with documented history of coronary, cerebral or peripheral vascular disease, where the functional status remains independent for activities of daily living and life expectancy more than 5 years 42,46,48,53,56,57 ; and (iii) statin therapy in diabetes mellitus if fasting serum cholesterol >50 mm or additional cardiovascular risk factor(s) present. 46,47,53 Several authors reported the percentage or number of criteria that were useful in identifying PIPs in their study populations. 17,46,53,54,56,57,60 For example, Gallagher and colleagues found that 861% of the STOPP and 100% of the START criteria were used in the identification of PIP in their study population. 46 Overall, in those studies using full, unmodified criteria, 17,46,54,56,57 431% 57 to 861% 46 of the 65 STOPP (median 569%) and 682% 57 to 100% 46 (median 887%) of the 22 START criteria were used in the identification of PIPs. Application of STOPP/START There were few challenges to applying either STOPP or START criteria documented in the thirteen studies. Five studies reported the need to consult multiple sources of medical history documentation to apply the criteria, including electronic and paper sources of information and direct contact with, caregivers and healthcare professionals. 42,53,54,55,57 Specifically, authors listed the following sources of medical history documentation: referral letters from general practitioners, own medications lists, pharmacy records, hospital admission records, chart reviews, telephone consultation with general practitioners and dispensing pharmacists, patient and/or caregiver interviews, prescription claims databases and electronic medical records. 42,53,54,55,57 One researcher noted that the inadequate documentation in charts of the rationale for prescribing or not prescribing medications made the application of the criteria difficult. 48 One researcher noted that there was a short but significant learning curve with the criteria 57, which may lead to higher application times until familiarity develops. Not all studies reported the training or background of the individuals applying the criteria (see Table 1). There was also little or no information available regarding the impact of the professional role, beliefs, intentions or goals of the researcher on the application of the criteria. Seven of the studies applied the criteria at a time when the patient was transitioning from one care setting to another. 42,46,48,52,54 56 As are often under stress with regard to their health when in the emergency department or when moving from one healthcare setting to another, it may be difficult to obtain an accurate medication history. In addition, the healthcare system may not facilitate accurate and timely medication reconciliation during care setting transitions. Four studies reported the time taken to apply the STOPP and/ or START criteria once the data were collected or extracted. 42,

7 An average of 3 min was taken to apply the START criteria in one study, 42 and 3 45 min was taken to apply both criteria Predictors of potentially inappropriate prescribing Predictors of PIP were reported in nine studies. 17,42,44,46,47,52,54,57,60 Age 75 years and older 44,47 or 85 years and older 42, female sex 42,44,54 and polypharmacy (multiple medications) 46,47,54 were all reported to be associated with increased odds of PIP. However, in one study, after adjusting for polypharmacy, the direction of the association between PIP and female gender and age was reversed. 44 Additionally, in one study, after adjusting for age, sex, comorbidity, chronic cognitive impairment, baseline activities of daily living function and number of medications, the likelihood of an ADE increased for each STOPP PIP instance. 52 Correlations were also observed between the number of medications prescribed, 17,57,60 increasing comorbidity as measured by the Charlson Comorbidity Index 57 and the identification of PIP by the STOPP criteria. Age 85 years and older, female sex, increasing comorbidity (Charlson Comorbidity Index greater than or equal to 2) and polypharmacy (10 or greater prescriptions) were also reported to be significantly associated with higher PPO (identified by START). 46 Heterogeneity in the break points used when converting continuous variables (age, polypharmacy (number of medications) and comorbidity (vs. multiple morbidity) into categorical variables make comparison of predictors of PIP and PPO difficult. Comparator explicit criteria Six studies compared the applicability and sensitivity of STOPP and the 2002 version of the Beers criteria. 17,46,52 54,57 One study 56 compared STOPP/START criteria to a set of Australian criteria. 28 All studies that compared the STOPP and Beers criteria found that STOPP was more sensitive than Beers. 17,46,52 54,57 In all of these studies, the Beers criteria were used as the standard PIP screening tools in older adults. The Beers criteria do not contain criteria for evaluating potential PPOs; however, a comparison between START and the Australian criteria developed by Basger and colleagues 28 provided a comparator for PPO detection. Results from this study show that the Basger s criteria appear to exhibit an increased sensitivity for PPO in older adults in longterm care. 56 Clinical and humanistic outcomes In a RCT, carried out by Gallagher and colleagues, it was reported that the application of the STOPP/START criteria in the form of an intervention, whereby a clinician recommends changes to the clinical team looking after the patient, significantly improved prescribing appropriateness. This was assessed by the Medication Appropriateness Index (absolute risk reduction of 357%) and the Assessment of Underutilization index (absolute risk reduction 212%). 55 This improvement was sustained over a 6-month followup period. The prevalence of falls and all-cause mortality was lower in the intervention group, but differences were not statistically significant (58% of the intervention group and 84% of the control group had at least one fall, P = 0332; 53% of the intervention group and 73% of the control group died, P = 0414). There was also a trend towards a lower frequency of primary care visits during the follow-up period in the STOPP/ START group (P = 0063). Two studies attempted to establish an association between the identification of PIP by the STOPP criteria and potentially avoidable ADEs. 52,54 Both studies found that the STOPP criteria was significantly more sensitive at identifying medication that may be associated with an ADE than with the Beers criteria. In the more recent study, each PIP as identified by STOPP increased the risk of ADEs by 847% (OR = 1847, 95% CI ; P < 0001); in contrast, PIPs identified by 2002 Beers criteria did not significantly increase the risk of an ADE (OR = 1276, 95% CI: ; P = 011). 52 Research involving the application of STOPP/START and the impact on the quality of life was not reported in any of the studies included in this review. Economic outcomes The direct costs of PIP were documented in three studies, two of which were performed in the Republic of Ireland 42,44 and one was performed in both Northern Ireland and the Republic of Ireland. 17 Barry and colleagues determined that the wholesale cost of the PPO instances identified by the START criteria in their study population was 188 per patient per year in Also, in 2007, Cahir and colleagues reported that the cost associated with the PIP instances defined by a condensed version of the STOPP criteria and identified in their study population was 318 per patient per year. 44 Byrne and colleagues determined that the cost associated with the PIP instances identified in their study population was 263 per patient per year (excluding prn medicines). 17 To date, no studies were found that performed a more detailed economic analysis relating to the cost-effectiveness of the intervention in terms of clinical or quality of life outcomes. (Table 2). Bias appraisal Critical appraisal for study bias was performed. Four studies were judged to have a low risk of bias, 46,54,55,57 two were rated as having moderate risk of bias 17,56 and two were rated as having high risk of bias. 52,53 We judged that seven studies adequately controlled for bias related to the study participation, outcome, application of STOPP/ START and confounding measurement domains. 17,46,52 55,57 We scored three as having a low risk of bias due to methods of data collection, 46,54,57 and five studies scored a low risk of bias due to approach for application of the STOPP/START tool. 17,46,54 56 One study was found to have a high risk of bias with regard to the application of the screening tool. 53 A moderate or high risk of bias was found among five of the eight assessed studies in the statistical analysis and data presentation domains. 17,46,52,53,57 Funding Reported funding was entirely by governmental or non-governmental organizations. The Health Research Board of Ireland supplied funding for six of the studies. 42,44,46,52,54,55 Additionally, Enterprise Ireland funding assisted Hamilton and colleagues, 52 and the Czech Ministry of Health Internal Grant Agency assisted Gallagher and colleagues with their international study. 46 Byrne and colleagues project was funded by the Centre for Ageing Research and Development in Ireland (CARDI). 17 The National Institute for Health Research assisted those working with Parsons and Johnston. 60 Pyszka reported no external funding or sponsorship, 48 and the remaining four studies did not report any source of funding. 47,53,56,57 366

8 Table 2. Outcomes of studies included in systematic review of STOPP/START criteria Source Criteria applied No. of pts Prevalence Most freq criteria applied Barry Full START % had 1 or et al. (2007) 42 more PPOs START: statins/ vascular disease; Warfarin/chronic atrial fibrillation; antiplatelet tx/arterial disease Gallagher & O Mahony (2008) 54 Full STOPP % had 1 or more PIMs STOPP: long-term BZDs; duplicate drug class; BZDs/history of falls Ryan Full STOPP/ et al. (2009) 57 START % had 1 or more PIMs, 227% had 1 or more PPOs STOPP: PPI at max tx > 8 wks; long-term BZDs; NSAIDs/ hypertension. START: antiplatelet tx/arterial disease; vitamin D/ osteoporosis; statins/ vascular disease Pyszka et al. (2010) 48 Modified STOPP/ Full START % had 1 or more PIMs, 468% had 1 or more PPOs on admission STOPP: medications without appropriate diagnosis (added criteria); ASA dosages >150 mg daily. START: statins/ vascular disease, ACE inhibitor/chf; vitamin D/ osteoporosis Conejos et al. Modified STOPP/ (2010) 53 START % had 1 or more PIMs, 43% had 1 or more PPOs STOPP: BZDs/history of falls; duplicate drug class. START: statins/vascular disease; statins/dm with cardiovascular risk Cahir et al. Partial STOPP % had 1 or more (2010) 44 PIPs STOPP: PPI at max tx > 8 wks; NSAIDS for > 3 consecutive months; long-term BZDs; duplicate drug class % Criteria detecting PIMs/PPOs 60% STOPP detected PIMs 431% STOPP detected PIMs, 682% STARTdetected PPOs 51% STOPP detected PIMs, 82% START detected PPOs Predictors of PIP: OR (95% CI), or Spearman s Rho (rs) Age 85: 208 ( ), P < 001 female sex: 229 ( ), P < 001 Female sex: 187 ( ), P = or less medications: 059 ( ), P = 0032 Number of medications prescribed: rs = 0356, P < 001; Age: rs = 0071, P < 001, comorbidity: rs = 0210, P < 001 Female sex: 110 ( ), 75 years a ;128 ( ) Clinic, humanistic, economic outcomes Other outcomes Economic: cost to supply PPO 3 min to apply START Hospitalization due to ADE 3 min to apply STOPP; 1 min to apply START Economic: cost of PIMs 367

9 Table 2 (continued) Source Criteria applied No. of pts Prevalence Most freq criteria applied % Criteria detecting PIMs/PPOs Predictors of PIP: OR (95% CI), or Spearman s Rho (rs) Clinic, humanistic, economic outcomes Other outcomes Liu et al. Full STOPP/ (2011) 47 START % had 1 or more PIMs. 419% had 1 or more PPOs STOPP: BZDs, neuroleptics, 1st generation antihistamines/ history of falls; Ca channel blockers/ chronic constipation. START: statins/dm with cardiovascular risk; antiplatelet tx/ arterial disease; metformin/type 2 DM Age 75-84: 190 ( ), P = 0011; age 85:227 ( ), P = 0006; 5 8 medications 266 ( ), P = 0001, 9 medications 718 ( ), P = 0001 Gallagher et al. (2011) 55 Full STOPP/ START 382 Mortality, GP visits, hospital readmissions, and falls Medication Appropriateness Index (MAI) and Assessment of Underutilization (AOU) index at multiple times over 6-month period, 3 min to apply STOPP/ START Gallagher et al. (2011) 46 Full STOPP/ START % (347 to 773% by country) had 1 or more PIMs, 594% (513 to 727% by country) had 1 or more PPOs STOPP: BZDs, neuroleptics/history of falls; duplicate drug class; PPI at max tx >8 weeks. START: vitamin D/ osteoporosis; statins/ vascular disease; statins/dm with cardiovascular risk 861% STOPP detected PIMs, 100% START detected PPOs STOPP: 6-10 medications: 231 ( ), P < 0001, >10 medications: 722 ( ), P < 0001, P < 0001; START: Age 85: 180 ( ), P = 0006; CCI 2: 325 ( ), P < 0001; dementia: 070 ( ), P =

10 Table 2 (continued) Source Criteria applied No. of pts Prevalence Most freq criteria applied % Criteria detecting PIMs/PPOs Predictors of PIP: OR (95% CI), or Spearman s Rho (rs) Clinic, humanistic, economic outcomes Other outcomes Byrne et al. (2011) 17 Full STOPP 630 RoI: 73% had 1 or more PIMs. NI: 67% had 1 or more PIMs Garcõa-Gollarte Full STOPP/ et al. (2011) 56 START % had 1 or more PIMs. 74% had 1 or more PPOs STOPP: PPI at max tx >8 weeks; BZDs/ history of falls; duplicate drug class; neuroleptics/history of falls STOPP: PPI at max tx > 8 weeks; BZDs/ history of falls; neuroleptics/history of falls. START: vitamin D/ osteoporosis; statins/ vascular disease; antiplatelet tx/arterial disease 69% STOPP detected PIMs 569% STOPP detected PIMs, 954% START detected PPOs RoI: total number of medications: rs = 0356, P < 001; >5 medications rs = 0135, P < 001) NI: total number of medications: rs = 0356, P < 001; >5 medications rs = 0256, P < 001) Economic: costs of PIMs 3 min to apply STOPP/START Hamilton et al. Full STOPP % had 1 or (2011) 52 more PIMs STOPP: PPI at max tx > 8 weeks; ASA/no history of vascular disease; BZDs/ history of falls; duplicate drug class Composite of age, sex, comorbidity, chronic cognitive impairment, baseline activities of daily living function, number of medications: 185 ( ), P < 0001 Adverse drug events Parsons et al. Partial STOPP % and 409% (2012) 60 had 1 or more PIMs at 2 time points STOPP: long-term neuroleptics; NSAIDS >3 weeks; PPI at max tx >8 weeks 20% STOPP detected PIMs Number of medications: rs = 0335, P < 001 STOPP, screening tool of older person s prescriptions; START, screening tool to alert doctors to right treatment; PPO, potential prescribing omissions; PIM, potentially inappropriate medication; PIP, Potentially inappropriate prescribing; BZD, benzodiazepines; PPI, proton pump inhibitors; ACE, angiotensin-converting enzyme; NSAID, non-steroidal anti-inflammatory drug; ASA, acetylsalicylic acid; Ca, calcium; DM, diabetes mellitus; CHF, congestive heart failure; CCI, Charlson Comorbidity Index; tx, treatment; RoI, Republic of Ireland; NI, Northern Ireland; RCT, randomized controlled trial. a Adjusted for gender. 369

11 Discussion The prevalence and predictors of PIP reported in this review may provide useful benchmarks for prescribing quality in older adults in the future. The generalizability of these findings may be limited by the marked variation and heterogeneity in the study design, PIP detection methodologies employed and study populations between the different studies. An association between the application of STOPP criteria and the identification of potentially avoidable ADEs has been reported, 52,54 and this may lead to an impact on clinical outcomes. Gallagher and colleagues provide evidence of improved and sustained prescribing appropriateness when STOPP and START criteria are applied. 55 However, as acknowledged by the authors, the study was not powered to detect a clinically significant difference between groups in regard to all-cause mortality, nor was the follow-up sufficiently long to allow detection of a potentially significant reduction in the prevalence of falls or readmission. 55 Overall, the evidence is not sufficiently robust to determine the impact that the STOPP/START criteria may have on clinical or humanistic outcomes. Research on the use of STOPP/START appears to still be in its infancy, but it is gaining momentum. With agencies such as the European Union Geriatric Medicine Society providing their backing to these criteria, 51 it is likely that they will continue to be used. Future research should employ strong methodological design and be sufficiently powered. It will be interesting to see how the 2012 version of Beers criteria fares in comparative studies. These future studies should not only focus on PIP, but should also examine the relationship between PIP and patient-related outcomes. Future research could also focus on the development and evaluation of intervention strategies to improve prescribing. There were limitations associated with this review at the study level, between studies and at the review level. With the exception of one publication 55, the STOPP/START studies used an observational design, five of which used retrospective data. 44,47,48,53,60 For the prospectivestudies,astrategyofconsecutivesamplingwasemployed in an additional six studies. 42,46,52,54,56,57 Several studies had sample populations of 150 or less, 48,53,56,60 and several studies had populations that were restricted to one hospital. 42,47,48,52,54 The criteria applied varied between studies. Although referred to as STOPP or START, some researchers used versions of the criteria that had been modified for their jurisdictional prescribing practices or formularies. 48,53 The criteria were also shortened in two studies to apply the criteria to certain healthcare databases not holding or linked to clinical data such as diagnoses and laboratory results. 44,60 Not all researchers had access to complete medication profiles including over-the-counter medications. One study reported that they had access to medication lists from the hospital studied, but that may have simultaneously been taking other medications prescribed by other physicians. 47 No study indicated an attempt to document or evaluate adherence to drug therapy. Researchers who had used pharmacy claims data were only able to confirm that had made a claim for medications, not that they have actually taken the medication. These inconsistencies may limit the generalizability of the results. We were unable to assess studies published in languages other than English. 45,61 64 Furthermore, at least 20 studies were only available as abstracts and were excluded. Following the cut-off of our systematic review period (January 2012), researchers from the United Kingdom, Ireland, Malaysia, Spain, Belgium, Germany and Australia have published primary studies using the STOPP/START criteria. 12,65 74 This work expands the number of jurisdictions researching and publishing on the criteria and increases the experience of the criteria in long-term care facilities. Two publications may warrant additional attention: Dalleur and colleagues established a relationship between STOPPidentified PIPs and hospital admissions, 69 and Ryan and colleagues compared the application of the STOPP/START criteria with and without clinical data. 72 This latter research has implications for those wishing to apply the criteria to digital databases. The majority of the work continues to evaluate the prevalence and predictors of PIP, however, and does not significantly change the results of our review. There were no comparisons between STOPP/START and the 2012 iteration of Beers criteria found. WHAT IS NEW AND CONCLUSION The STOPP/START criteria provide a promising framework for assessment of PIP in older adults. They may be more applicable than other explicit criteria for the assessment of PIP across all care settings from community to long-term care. Further work is needed to evaluate the true transferability of this set of criteria across the different care settings and jurisdictions. Research using the STOPP/START criteria has documented the prevalence of PIP in older adults in multiple healthcare settings and jurisdictions. In two observational studies, there also appears to be an association between the identification of PIP by STOPP and potentially avoidable ADEs; however, the true extent of this association still has to be elucidated. To date, the clinical, humanistic and economic impacts of the application of the STOPP/START criteria have not been well explored. FUNDING This review was funded, in part, by the Drug Evaluation Alliance of Nova Scotia. Dr. Ingrid Sketris held a chair funded by CHSRF/ CIHR and co-sponsored by NHSRF. Barbara Hill-Taylor received funding from this chair. David O Sullivan received funding from the Health Research Board of Ireland grant no: HRA_HSR/2010/ 14 Dr. Stephen Byrne would like to acknowledge the support of The Ireland Canada University Foundation- Dobbin Scholarship. REFERENCES 1. United Nations, Department of Economic and Social Affairs, Population Division. World Population Ageing. Document ESA/P/WP/ Available at: WPA2009/WPA2009_WorkingPaper.pdf (accessed 22 November 2012). 2. Canadian Institutes of Health Research. Canadian Longitudinal Study on Aging Available at: e/18542.html (accessed 22 November 2012). 3. Cruz-Jentoft A, Boland B, Rexach L. Drug therapy optimization at the end of life. Drugs Aging, 2012;29: Hanlon JT, Schmader KE, Ruby CM, Weinberger M. Suboptimal prescribing in older in and out. J Am Geriatr Soc, 2001;49: Reason B, Terner M, McKeag AM, Tipper B, Webster G. The impact of polypharmacy on the health of Canadian seniors. Fam Pract, 2012;29: O Mahony D, Gallagher PF. Inappropriate prescribing in the older population; need for new criteria. Age Ageing, 2008;37: