Effectiveness of Interventions to Reduce Potentially Inappropriate Medication in Older Patients: A Systematic Review

Background: Age-related multiple comorbidities cause older adults to be prone to the use of potentially inappropriate medicines (PIM) resulting in an increased risk of adverse events. Several strategies have emerged to support PIM prescription, and a huge number of interventions to reduce PIM have been proposed. This work aims to analyze the effectiveness of PIM interventions directed to older adults. Methods: A systematic review was performed searching the literature in the MEDLINE PubMed, EMBASE, and Cochrane scientific databases for interventional studies that assessed the PIM interventions in older adults (≥65 years). Results: Forty-seven articles were included, involving 52 to 124,802 patients. Various types of interventions were analyzed such as medication review, educational strategies, clinical decision support system, and organizational and multifaceted approaches. In the hospital, the most successful intervention was medication review (75.0%), while in primary care, the analysis of all included studies revealed that educational strategies were the most effective. However, the analysis of interventions that have greater evidence by its design was inconclusive. Conclusion: The results obtained in this work suggested that PIM-setting-directed interventions should be developed to promote the wellbeing of the patients through PIM reduction. Although the data obtained suggested that medication review was the most assertive strategy to decrease the number of PIM in the hospital setting, more studies are necessary. Systematic Review Registration: [https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42021233484], identifier [PROSPERO 2021 CRD42021233484].


INTRODUCTION
The increase in life expectancy associated with a declined birth rate contributed to rapid population aging (United Nations, 2019). Even though the world population is getting older, aging populations differ by region and level of development (Beard et al., 2016). Globally, it is estimated that in 2050 the number of older adults will reach 1.5 billion and will outnumber adolescents and youth aged 15-24 years (1.3 billion) (United Nations, 2019).
Considering that more than half of older adults have at least two chronic diseases (Barnett et al., 2012), these societal transformations pose a significant challenge in health systems and increase the consumption of health resources, including medicines. In addition, the treatment of chronic diseases is based on single disease-centered guidelines that can lead to an overwhelming of medication, and polypharmacy can easily occur (Barnett et al., 2012). Age-related pharmacokinetic and pharmacodynamic alterations associated with the use of multiple medicines can potentiate the consumption of potentially inappropriate medications (PIM) and facilitate the occurrence of adverse drug reactions (ADR) in frail older adults (Motter et al., 2018;Hefner et al., 2021).
PIM is defined as medicines that should not be prescribed because the risk of adverse events outweighs the clinical benefit, especially when more effective alternatives are available (Renom-Guiteras et al., 2015). The prescription of PIM has received special attention from the health community, and interventions aim to optimize medication prescribing and increase the benefit/risk ratio associated with the patients (Anderson et al., 1997;Simonson and Feinberg, 2005). In the last decades, several studies have been done to evaluate the effectiveness of PIM interventions in primary care, such as in hospitals and nursing homes. Nevertheless, the studies display widely differing methodology and inconsistent results, and to our knowledge, there are no systematic reviews comparing the effectiveness of different kinds of interventions.
Thus, this study aims to critically review the effectiveness of interventions to reduce PIM prescriptions in older adults.

METHODS
This systematic review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines (Page et al., 2021) Table S1). The research protocol is registered on PROSPERO (CRD42021233484).

Search Strategy
A literature search was conducted in January 2021 and updated in February 2021 on the MEDLINE PubMed, EMBASE. A search was also conducted in the Cochrane database in October 2021. The search strategy was designed to identify relevant studies addressing interventions on PIM prescriptions in older adults, using the following broad-based search terms strategy: "(elderly OR "elderly patient" OR "older patient*" OR "older adult*" OR "geriatric patient*") AND (PIM OR PIP OR "potentially inappropriate medicine" OR "potentially inappropriate medication" OR "potentially inappropriate prescribing" OR "prescribing patterns" OR "Prescription Drug Misuse" OR "Prescription Drug Overuse" OR "deprescri*" OR "potentially inappropriate prescription*") AND (prevention OR reduction OR decrease OR impact) AND (intervention OR trial)."

Selection Criteria
This systematic review included the following: 1) all studies focused on PIM interventions directed to older adults (≥65 years) that aimed to optimize their pharmacotherapy; 2) controlled intervention studies and case series studies; and 3) all studies published in Portuguese, English, or Spanish between January 1, 2000, andDecember 31, 2020. Excluded from this work were reviews, meta-analyses, opinions, letters to the editor that do not provide original data, comments, reports, studies addressing PIM in a specific pathology, and studies targeting a limited and predefined class of PIM.

Outcomes Measures
Our primary outcome measure was the effectiveness of the PIM interventions through the analysis of the change rate between the mean number of PIM per patient and/or the mean number of patients with PIM before and after an intervention.

Data Extraction
Two researchers (AP and DR) independently screened all titles and abstracts retrieved from the databases accordingly with the inclusion criteria. To evaluate the eligibility of full-text articles, two researchers (AP and DR) independently screened the full text of the articles. All discrepancies were resolved through discussion with the help of a third researcher (FR).

Quality Assessment
Two researchers (AP and DR) independently evaluated the quality and susceptibility to the bias of the included studies using the "Quality Assessment of Controlled Intervention Studies" and the "Quality Assessment Tool for Case Series Studies" tools, depending on study design (National Heart Laboratory, 2013). All discrepancies were resolved through discussion with a third (FR) or fourth researcher (MH).

Data Synthesis and Presentation
Two researchers (AP and DR) independently extracted data from the included studies. The data extracted from each article include authors, publication year, study design, country, sample size, patients' age, type of intervention applied, PIM screening tool, outcome measures, and main results.
To better analyze the extracted data, studies were grouped according to the intervention used. Within the intervention used, studies were grouped according to the setting where the intervention occurs. Five different interventions were identified in the included studies, and their descriptions were based on the following pre-defined definitions: 1) Medication review: "a structured evaluation of a patient's medicines to optimize medicines use and improve health outcomes. This entails detecting drug-related problems and recommending interventions" (Griese-Mammen et al., 2018); 2) educational interventions: "a package of interventions aimed to refresh the basic pharmacology competencies of a healthcare professional to change the prescription. The approaches used in the educational interventions included: interactive teaching, mailed educational material combined with individual feedback, and face-to-face visits to physicians" (Kaur et al., 2009); 3) clinical decision support systems (CDSS): "electronic tools that prompt provider behaviors in various areas of patient care, including medication ordering, chronic disease management, health care screening, and vaccination. CDSS can provide physicians, nurses, pharmacists, and other care providers with patient-specific prompts or warnings, treatment guidelines (e.g., order sets), automatic medication dosing calculators, or reports of overdue tests and medications as appropriate" (Bhugra and Cutter, 2001); 4) multifaceted interventions: "any intervention including two or more components." In this study we classified as a multifaceted approach studies that used a combination of the interventions described above (Squires et al., 2014); and 4) organizational strategies: a combination of methodologies to improve the quality indicators. This type of intervention can include several methodologies, such as diagnostic activity, team building, intergroup relationship, sensitivity training, etc. In this work, organizational strategies include showing charts with the percentage of patients with PIM, an educational session with practices to identify patients with PIM, and frequent reunions to evaluate PIM indicators (Cady and Kim, 2017).

Statistical Analysis
assessed through the comparison between the interventions. The efficacy of the interventions was presented as a change rate between the mean number of PIM per patient and/or the mean number of patients with PIM before and after an intervention.

Study Selection
The search of the databases yielded 3,406 citations ( Figure 1). After screening titles and abstracts, 98 articles potentially met the inclusion criteria. Because seven articles were not retrieved, only 91 articles were fully screened. Among these, 42 were excluded because they did not address interventions (n 6), did not report PIM specific outcomes (n 15), addressed PIM in patientspecific diseases (n 4), addressed a pre-selected and/or a limited number of medicines (n 14), and did not address older patients (≥65 years old) (n 5) (Supplementary Table S2).

PIM Screening Tools
Thirty-seven studies used validated and published criteria to identify PIM, including Beers criteria (n 16) (Brown and Frontiers in Pharmacology | www.frontiersin.org January 2022 | Volume 12 | Article 777655    A, after; B, before; C, control group; I, intervention group; IQR, interquartile range; RCT, randomized controlled trial; SD, standard deviation. a The National Institutes of Health (NIH) quality assessment tool for case series studies. b The National Institutes of Health (NIH) quality assessment tool of controlled intervention study. c Median age. Was the method of randomization adequate (i.e., use of randomly generated assignment)? 11 2 9 3 Was the treatment allocation concealed (so that assignments could not be predicted)? 10 3 9 4 Were study participants and providers blinded to treatment group assignment? 6 11 5 5 Were the people assessing the outcomes blinded to the participants' group assignments? 8 8 6 6 Were the groups similar at baseline on important characteristics that could affect outcomes (e.g., demographics, risk factors, co-morbid conditions)?
19 2 1 7 Was the overall drop-out rate from the study at endpoint 20% or lower of the number allocated to treatment? 13 9 0 8 Was the differential drop-out rate (between treatment groups) at endpoint 15 percentage points or lower? 19 1 2 9 Was there high adherence to the intervention protocols for each treatment group? 22 0 0 10 Were other interventions avoided or similar in the groups (e.g., similar background treatments)? 21 1 0 11 Were outcomes assessed using valid and reliable measures, implemented consistently across all study participants? 22 0 0 12 Did the authors report that the sample size was sufficiently large to be able to detect a difference in the main outcome between groups with at least 80% power? 12 4 6 13 Were outcomes reported or subgroups analyzed prespecified (i.e., identified before analyses were conducted)? 13 2 7 14 Were all randomized participants analyzed in the group to which they were originally assigned, i.e., did they use an intentionto-treat analysis?     (Tallon et al., 2015). Five studies used self-developed or adapted criteria (Allard et al., 2001;Keith et al., 2013;Lopatto et al., 2014;Van der Linden et al., 2017;Van Der Linden et al., 2018), and the remaining studies used a combination of validated (STOPP and/or Beers) and self-developed or adapted criteria.
Six (Brown and Earnhart, 2004;Spinewine et al., 2007;Tallon et al., 2015  in the average number of PIM after an intervention. Finally, Brown and Earnhart (2004) conclude that their intervention led to an 8.08% absolute risk reduction. Among the studies performed by physicians in hospitals, one was conducted at an emergency department of a hospital (Gallagher     (Keith et al., 2013;Lopatto et al., 2014). Gallagher et al. (2011) observed a significant reduction in the proportion of patients with at least one PIM in the intervention group (from 43.2% in admission to 3.7%, at discharge). This trend remains stable during the 6 months of follow-up. One study performed a quality improvement program across an Italian region with more than 80,000 older adults and observed that the PIM exposure incidence rate significantly declined 31.4% (from the baseline to the postintervention period) (Keith et al., 2013). The intervention performed by a gerontologist in hospitalized patients results in a significant decrease in the number of PIM in patients discharged (PIM discontinuation of 39.7%) (Dalleur et al., 2014). Similar outcomes were achieved in the intervention performed by investigators in hospitalized patients (Van Der Linden et al., 2018).
In primary care, the medication review performed by both physicians and pharmacists results in a significant reduction in the mean number of PIM per patient [from 0.6 to 0.4 (Hurmuz et al., 2018) and from 1.24 to 0.82 (Stuckey et al., 2018)]. In this setting, a multidisciplinary team failed to achieve a significant reduction in the number of PIM, and only one (Khera et al., 2019) of the three pharmacists' interventions (Castelino et al., 2010;Campins et al., 2016;Khera et al., 2019) results in a significant impact in PIM reduction.
Finally, in a chronic care geriatric facility, pharmacy students observed a decline in PIM prescriptions after a follow-up of 12 and 24 months (Frankenthal et al., 2014;Frankenthal et al., 2017).

Impact of Educational Interventions
In the eight included studies (Fick et al., 2004;Rognstad et al., 2013;Franchi et al., 2014Franchi et al., , 2016Ilić et al., 2015;Clyne et al., 2016;Etxeberria et al., 2018;Moss et al., 2019) that used educational approaches to reduce PIM, the interventions were performed by the following: a researcher and/or a research team (Fick et al., 2004;Franchi et al., 2014Franchi et al., , 2016Ilić et al., 2015;Etxeberria et al., 2018), a multifaceted team containing a pharmacist and a physician (Moss et al., 2019) or physicians (peer academic detailers) (Rognstad et al., 2013), or a pharmacist (Clyne et al., 2016). In all studies the target of the educational interventions were physicians. The outcomes of the interventions were measured through the reduction of PIM use or PIM prescriptions. In one of the studies, the included population was polymedicated older adults (Etxeberria et al., 2018); in the three studies that have a positive impact on PIM, the average number of PIM per patient ranged from 0.7-11 before intervention to 0.51-1.5 after the intervention (Ilić et al., 2015;Clyne et al., 2016;Etxeberria et al., 2018). One study reported that the number of physicians that prescribed at least one PIM declined 17.9% (Fick et al., 2004). Finally, one study reported that after the educational intervention the PIM rate ratio before and after the intervention is 0.73 (Moss et al., 2019).

Impact of Clinical Decision Support System Interventions
Five studies used CDSS to reduce PIM (Urfer et al., 2016;Price et al., 2017;Vanderman et al., 2017;McDonald et al., 2019;Winata et al., 2020). Two of four studies performed in hospitalized patients reported that the implementation of CDSS has a positive impact on PIM deprescription (Urfer et al., 2016;Vanderman et al., 2017). In one study, the introduction of a PIM checklist, in an internal medicine ward, leads to a significant reduction (22.0%) of the risk of  being prescribed one or more PIM (Urfer et al., 2016). Finally, one study observed that although the total number of newly prescribed PIM did not decrease, the top 10 most common new PIM significantly decreased from 9.0% to 8.3% (Vanderman et al., 2017).

Impact of Multifaceted Interventions
Nine studies used a multifaceted approach as a strategy to decrease the number of PIM (Clyne et al., 2016;Moss et al., 2016;Gibert et al., 2018;Najjar et al., 2018;Vandenberg et al., 2018;Boersma et al., 2019;Liu et al., 2019;Vu and Huong, 2019;  Akkawi et al., 2020) In two studies, the multifaceted approach consisted in the application of a CDSS followed by a medication review (Boersma et al., 2019;Liu et al., 2019). One study used a combination of educational and CDSS approaches (Akkawi et al., 2020). The remaining studies used, as an approach to decrease the number of PIM, a combination of an educational approach followed by a medication review (Clyne et al., 2016;Moss et al., 2016;Gibert et al., 2018;Najjar et al., 2018;Vandenberg et al., 2018;Vu and Huong, 2019).

Clinical Decision Support System and Medication Review
The combined use of CDSS and medication review strategies led to a significant reduction of the mean number of PIM per patient from 0.7 to 0.23 (Boersma et al., 2019) and an increasing number of PIM changes implementation (intervention: 46.2% vs. control: 15.3%) (Liu et al., 2019).

Educational Intervention and Medication Review
Five of the six studies that used a combination of educational and medication review strategies observed a significant impact on the PIM reduction (Clyne et al., 2016;Gibert et al., 2018;Najjar et al., 2018;Vandenberg et al., 2018;Vu and Huong, 2019). Among these, two studies reported a significant decrease in the mean number of PIM per patient from 0.99-1.18 before intervention to 0.66 to 0.7, after intervention (Clyne et al., 2015;Gibert et al., 2018). Vandenberg et al. (2018) and Vu and Huong (2019) reported that the prevalence of PIM decreased 5.9% and 10.9%, respectively. According to Najjar et al. (2018), the multifaceted approach led to a decrease in the PIM incidence of 31.5%.

Educational Intervention and Clinical Decision Support System
In one study (Akkawi et al., 2020), the multifaceted intervention consists of three educational sessions about PIM and discussion of STOPP criteria, coupled with an introduction of a CDSS. This approach did not achieve significant outcomes.

Impact of Organizational Interventions
One study performed an organizational intervention in four different hospitals and observed a significant decrease in the average percentage of prescribed PIM per month (1.7-6.8) after intervention (Stevens et al., 2017). Another study performed in 99 primary care practices observed that the organizational intervention that includes PIM performance reports, on-site visits, and network meetings was responsible for an absolute annual decline of 0.018% for always inappropriate medication (Wessell et al., 2008).

Impact of Interventions That Have Greater Evidence by Its Design
The analysis of the studies that included a concurrent control and low risk of bias revealed that in the hospital, all the five medication review interventions were effective (Spinewine et al., 2007;Gallagher et al., 2011;Dalleur et al., 2014;Van der Linden et al., 2017;Van Der Linden et al., 2018), two CDSS interventions were effective (Urfer et al., 2016;McDonald et al., 2019), one multifaceted intervention achieved significant impact (Vandenberg et al., 2018), and none of the educational interventions achieved a successful reduction of PIM. In primary care it was observed that all the two multifaceted interventions achieved a significant reduction of PIM (Clyne et al., 2015;Boersma et al., 2019), one (Clyne et al., 2016) of two educational strategies was effective, and none of the medication review and CDSS strategies was well successful.

DISCUSSION
Despite the extensive number of studies in the literature on PIM in older patients, only 31 of the included studies reported effective intervention. Among these, 21 presented methodological intervention limitations and could not ensure that the intervention used can be replicated and identical outcomes achieved (Brown and Earnhart, 2004;Fick et al., 2004;Spinewine et al., 2007;Wessell et al., 2008;Gallagher et al.,  Organizational intervention-a combination of strategies to improve the quality indicators of institutions/organizations and enrolled in the approach all stakeholders, health professionals, and non-health professionals. This intervention uses several approaches, including diagnostic activity (including medication review), Team Building, Intergroup relationship, sensitivity training (including educational sessions. Frontiers in Pharmacology | www.frontiersin.org 2011; Dalleur et al., 2014;Frankenthal et al., 2014;Clyne et al., 2015;Tallon et al., 2015;Urfer et al., 2016;Frankenthal et al., 2017;Van der Linden et al., 2017;Etxeberria et al., 2018;Gibert et al., 2018;Hurmuz et al., 2018;Van Der Linden et al., 2018;Boersma et al., 2019;Gutiérrez-Valencia et al., 2019;Khera et al., 2019;Liu et al., 2019;Vu and Huong, 2019). Although a metaanalysis was not done, our findings suggested that in the hospital, the most adequate strategy to decrease the number of PIM and/or the patients with at least one PIM was medication review. Concerning primary care setting, the analysis of all the included studies indicated that educational interventions were the most successful. However, when only randomized controlled trial (RCT) studies were analyzed, it did not find greater effectiveness of some interventions over others. The data of this study also suggested that the inclusion of pharmacists can upgrade the quality of the PIM intervention and effectively promote the well-being of the patients.
Regarding the influence of the number of prescribed medicines per patient in PIM interventions, our data suggested that the success of an intervention is not medicines number-dependent, since the analysis of the successful intervention rate in polymedicated and non-polymedicated patients was similar (≈67%).
This work also suggested that most of the studies presented important design limitations, something that limits the grade of their evidence.
Medication review was the most frequent strategy used to improve pharmacotherapy and reduce the number of PIM in hospitalized patients. A reduction in the number of PIM per patient or/and in the number of patients with at least a PIM was achieved for 75% of the medication review interventions (Brown and Earnhart 2004;Spinewine et al., 2007;Gallagher et al., 2011;Keith et al., 2013;Dalleur et al., 2014;Ilić et al., 2015;Tallon et al., 2015;Van der Linden et al., 2017;Chan et al., 2018;Fajreldines et al. 2018;Van Der Linden et al., 2018;Gutiérrez-Valencia et al., 2019). Among the three studies that do not have efficacy in hospitalized patients, the main reasons pointed were as follows: 1) the difficulty to engage physicians to actively participate in the study-they preferred to receive the documentation about drug therapy issues by paper instead of discussing face-to-face the patients' pharmacotherapy (Chan et al., 2018); and 2) the low acceptance of the recommendation by the physicians (Regueiro et al., 2019).
In primary care, 42.9% of the interventional studies (Keith et al., 2013;Hurmuz et al., 2018;Stuckey et al., 2018;Khera et al., 2019) used medication review to improve the pharmacotherapy through the reduction of PIM. The lack of efficacy can be related to 1) the low acceptance rate of recommendations by the physicians (Allard et al., 2001); 2) the PIM list used-for example, Castelino et al. reported that in Australia the medicines listed in Beers criteria were rarely used; 3) the physicians did not use routinely any checklist and do not access computer programs to evaluate hypothetical interactions, and they do not record short-term drug alterations (Lampela et al., 2010); 4) in some cases the patients did not receive the full intervention (Allard et al., 2001); and 6) contamination between control and intervention groups (Campins et al., 2016).
The analysis of all educational interventions performed in primary care revealed that this type of intervention has been successfully implemented in 75%. However, only one of the two studies that have greater evidence by its design effectively decreased the number of PIM. The success of educational interventions in primary care can be related to the promotion of a specific web training on PIM tools used by physicians in clinical practice, updating the knowledge of physicians in PIM detection (Fick et al., 2004;Clyne et al., 2016;Etxeberria et al., 2018). The lack of efficacy of educational intervention observed in one study can be related to a change in the participants' behavior due to the knowledge that they are taking part in an experiment (Hawthorne effect) (Rognstad et al., 2013).
In hospitalized patients, the poor outcomes achieved by educational interventions can be related to the low interactivity during the education intervention, the lack of knowledge of the clinicians, and the characteristics of the ward included that sometimes make difficult the collection of the data (Franchi et al., 2016).
The implementation of a CDSS in hospitals had a positive impact on 50% of the studies (Urfer et al., 2016;Vanderman et al., 2017). The lack of efficacy of the intervention in the remaining studies can be related to the study design and the fact that the applied criteria are not setting-directed originating a high number of alerts that tend to be ignored by a healthcare professional (McDonald et al., 2019;Winata et al., 2020).
Multifaceted interventions were described as mixed interventions that can reduce the number of PIM (Rahme et al., 2005). In the hospital setting, it was observed that in the two studies that used a combination of educational and medication review strategies, the intervention was well successful (Najjar et al., 2018;Vu and Huong, 2019). In primary care, a combination of educational and medication review strategies results in increased efficacy of the intervention (Clyne et al., 2015;Gibert et al., 2018;Vandenberg et al., 2018).
The results of the included studies suggested that medication review is the most indicated intervention to promote the wellbeing of the hospitalized patients through the reduction of PIM. The success of medication review strategies at hospital discharge could be related to the fact that the inpatient setting may predispose older adults to new prescriptions and probably unnecessary drugs (Page et al., 2010). Moreover, during the hospitalization physicians tend to resist the change or discontinuation of chronic medication, particularly if the medication is not related to the reason for hospitalization (Page et al., 2010). This high number of prescribed PIM during the hospitalization can be the result of the lack of implemented PIM programs directed to each hospital ward and/or specific condition (Motter et al., 2018).
To improve the well-being of older adults, besides strategies to reduce PIM, strategies to promote appropriate prescription have also been developed. Medication review is a widely used strategy and with better outcomes to reduce potentially inappropriate prescribing (PIP) in hospitalized older adults. However, in a recent review, Dautzenberg et al. (2021) reported that the heterogeneity between studies does not allow reaching significant conclusion. According to dos , the choice of outcome measures, study design, and methodological quality of medication review studies make it difficult to analyze the effectiveness of this strategy. The failure of medication review strategies in primary care can be attributed to the lack of time of physicians to perform the medication review (Plácido et al., 2020); also, as a result of this lack of time, even when the medication review was performed, patients' follow-up did not occur (Campins et al., 2016). On the other hand, educational strategies allow the empowerment of primary care physicians who already had enough handling in managing older adults but not the right confidence and knowledge to manage PIM prescription (Maio et al., 2011). Moreover, educational strategies had more impact on prescribing patterns than presenting a physician only with a decision algorithm (Rahme et al., 2005). A previous work observed similar results regarding the effectiveness of educational strategies to reduce PIP in primary care. According to Kunstler et al. (2019), educational strategies are well successful in changing health professional prescribing behavior.
A recent systematic review focused on non-clinical programs to reduce the inappropriate or unnecessary use of medicines observed that interventions consisting of education messages and recommended behavior alternatives were more likely to be successful in reducing the inappropriate use of medicines or medical procedures (Lin et al., 2020). Educational strategies are essential to improve prescription, as observed by Amorim et al. (2021) since physician-related characteristics can influence the number of PIM prescriptions.
Regarding the multifaceted strategies, the scarcity of studies using this approach did not allow clarifying their benefits in PIM reduction.
In the hospital setting, CDSS interventions significantly reduced the PIM number in older adults. Similar results were found in another systematic review (Dalton et al., 2018). The lack of CDSS in primary care can be related to the outdated user interface model (Price et al., 2017).
Regarding the organizational strategies, the studies achieved a significant impact on pharmacotherapy independently of the setting (Wessell et al., 2008;Stevens et al., 2017).
In 47.8% of the included studies, the intervention was performed by a pharmacist or by a multifaceted team that includes at least one pharmacist. Among these studies, the rate that interventions succeeded well was 72.7%. In the remaining studies, the rate of success observed was 62.5%, suggesting that the inclusion of a pharmacist in the PIM interventions team can be beneficial. Previously it was demonstrated that pharmacists are actively engaged in several care-delivery models such as direct patients care and collaborative team-based care, improving pharmacotherapy and ameliorating the patients-related health outcomes (Lee et al., 2015). It was also reported that pharmacists could play an important role in patients' medication review in practice settings such as community pharmacies long-term care facilities, outpatient clinic home care, and hospitals. Moreover, pharmacists-led deprescribing interventions can reduce the number of unnecessary and potentially harmful medications (Silva et al., 2019;Hernández-Prats et al., 2021).
Although this study was performed with scientific rigors, some limitations are present. The search strategy was limited to the three main health research databases and articles written in English, Portuguese, and Spanish. The included studies were heterogeneous in practice settings, population, size of the samples, and PIM definition that can be variable depending on the screening tool used, which can influence PIM number detected (Thomas and Thomas, 2019;Perpétuo et al., 2021).
Because this review includes studies independently of the quality assessment analysis, an outcomes bias can be aroused. The bias can be attributed to a lack of randomization and blinded interventions and absences/inadequate follow-up period in some studies, compromising a possible scaling up of the interventions.
This study provided valuable data regarding PIM-reduction strategies; however, most of the included studies presented limitations that restrain the extrapolation of the results and a lack of an economic evaluation. Only one study reported that in the intervention group, a significantly lower medication cost was achieved (Frankenthal et al., 2014).
A recent systematic review only found seven articles reporting the economic impact of PIM interventions and suggested that although limited, interventions to optimize medication may outweigh their implementation costs (Laberge et al., 2021).

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.