Global prevalence of polypharmacy and potentially inappropriate medication in older patients with dementia: a systematic review and meta-analysis

Background: Older patients with dementia always need multiple drugs due to comorbidities and cognitive impairment, further complicating drug treatment and increasing the risk of potentially inappropriate medication. The objective of our study is to estimate the global prevalence of polypharmacy and potentially inappropriate medication (PIM) and explore the factors of PIM for older patients with dementia. Methods: We searched PubMed, Embase (Ovid), and Web of Science databases to identify eligible studies from inception to 16 June 2023. We conducted a meta-analysis for observational studies reporting the prevalence of potentially inappropriate medication and polypharmacy in older patients with dementia using a random-effect model. The factors associated with PIM were meta-analyzed. Results: Overall, 62 eligible studies were included, of which 53 studies reported the prevalence of PIM and 28 studies reported the prevalence of polypharmacy. The pooled estimate of PIM and polypharmacy was 43% (95% CI 38–48) and 62% (95% CI 52–71), respectively. Sixteen studies referred to factors associated with PIM use, and 15 factors were further pooled. Polypharmacy (2.83, 95% CI 1.80–4.44), diabetes (1.31, 95% CI 1.04–1.65), heart failure (1.17, 95% CI 1.00–1.37), depression (1.45, 95% CI 1.14–1.88), history of cancer (1.20, 95% CI 1.09–1.32), hypertension (1.46, 95% CI 1.05–2.03), ischemic heart disease (1.55, 95% CI 0.77–3.12), any cardiovascular disease (1.11, 95% CI 1.06–1.17), vascular dementia (1.09, 95% CI 1.03–1.16), chronic obstructive pulmonary disease (1.39, 95% CI 1.13–1.72), and psychosis (1.91, 95% CI 1.04–3.53) are positively associated with PIM use. Conclusion: PIM and polypharmacy were highly prevalent in older patients with dementia. Among different regions, the pooled estimate of PIM use and polypharmacy varied widely. Increasing PIM in older patients with dementia was closely associated with polypharmacy. For other comorbidities such as heart failure and diabetes, prescribing should be cautioned.


Introduction
The statistics of epidemiology revealed that in 2019, there were 703 million individuals aged 65 years or above living in the world, and the number was expected to reach 1.5 billion by 2050 (He and Kinsella, 2020).The global population aging would further accelerate the increase in the geriatric population, which imposed significant demands on the healthcare system.Meanwhile, increased medication use is one of the important challenges (Chiatti et al., 2012;Johnell, 2015).
Polypharmacy is defined as the concurrent use of multiple drugs, generally taking five or more drugs (Masnoon et al., 2017;Rankin et al., 2018).The older population often suffered multiple diseases, and polypharmacy was insufficient for controlling or curing diseases.A cross-sectional study performed by Chandrasekhar reported that the prevalence of polypharmacy in 210 inpatients aged 65 years or above was up to 60% and that of hyperpolypharmacy (ten or more drugs) was 35.7% (Chandrasekhar et al., 2019).Moreover, aging-related alteration in pharmacokinetics and pharmacodynamics might lead to stronger drug effects and prolongation of drug action time (Shi et al., 2008).Thus, the management of adverse effects on multiple drugs and potential drug-drug interaction among the older population was rather complicated and challenging.
Dementia, a degenerative nervous system disorder, features irreversible decline in cognitive function (Prince et al., 2013).Patients diagnosed with dementia were especially sensitive to adverse effects of central nervous system (CNS) drugs (Bell et al., 2012).Communication disorder caused by cognitive impairment and concomitant mental symptoms would lead to more complicated drug use in patients with dementia (Johnell, 2015).Furthermore, compared with non-dementia, dementia was more likely to be accompanied by other chronic diseases, such as hypertension and diabetes, exposing a higher risk of polypharmacy (Clague et al., 2017).Banta et al. indicated older patients with dementia are more likely to have five or more current prescriptions (Banta, 2017).Therefore, we should attach great importance to drug medications for older patients with dementia.
Potentially inappropriate medication (PIM) is an important concept to assess the quality of drug use.The term was first proposed by the American Panel in 1991, defined as those drugs with potential risks outweighing the benefits (Renom-Guiteras et al., 2015).A large number of studies have demonstrated PIM was associated with drug-related problems and adverse outcomes, such as increasing risk of hospitalization and death and incurring extra medical expenditure (Hagstrom et al., 2015;Hyttinen et al., 2017;Murphy et al., 2020).In order to evaluate PIM use and avoid the occurrence of adverse events, several explicit tools based on expert census were developed.A systematic review showed a total of 46 screening lists of PIM in the world that were identified, covering four continents and 13 countries (Kaufmann et al., 2014).The most frequently used criteria were Beers criteria and STOPP/START criteria (American Geriatrics Society Beers Criteria ® Update Expert Panel, 2019;O'Mahony et al., 2018), including drugs that should be avoided for treating common systemic diseases in elderly patients, possible adverse reactions, drug-drug interactions, drug-disease interactions, and risky drug use based on the renal function level.PIM use among older people was prevalent, especially in frail patients with dementia who need long-term care (Kristensen et al., 2018).Due to the difference of medical habits in each country and screening tools, the prevalence of PIM in patients with dementia varied widely.In European countries, 60% of older patients with dementia had at least one PIM based on the European Union (7)-PIM list (Renom-Guiteras et al., 2018).In China, the prevalence of PIM was 39.43% evaluated using 2019 Beers criteria (Zhao et al., 2022).Thus, a pooled analysis is necessary to conduct for evaluating the PIM and polypharmacy in patients with dementia, further providing a reference for countries that have not yet carried out a relevant study about drug burden in older patients with dementia.
To date, several reviews about PIM use and polypharmacy in dementia have been published (Johnell, 2015;Disalvo et al., 2016;Redston et al., 2018;Hukins et al., 2019;Delgado et al., 2020), but limited to a specific type of dementia or specific population (such as community or inpatients), or only qualitatively described the prevalence of PIM use or polypharmacy.In the systematic review and meta-analysis, we first summarized the pooled estimate of PIM and polypharmacy in older patients with dementia (not including mild cognitive impairment) across different regions and explored the association between PIM and polypharmacy and reviewed other factors associated with PIM use.

Methods
The study protocol of this systematic review and meta-analysis was registered on PROSPERO (CRD42022368310).The study was conducted based on MOOSE (Meta-analysis of Observational Studies in Epidemiology) guidelines (Stroup et al., 2000) and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (Moher et al., 2015).

Search strategy
A comprehensive search of PubMed, Embase (Ovid), and Web of Science was performed from inception to 16 June 2023.The search strategy was using a combination of Medical Subject Headings (MeSH) and free text words.The specific search details in different databases are listed in Supplementary Table S1.In addition, we performed manual searching of selected published full-text reviews, identifying other potential relevant articles.

Selection criteria
Studies were included if they recruited older adults with dementia (≥65 years, or mean age ≥70 years), reported the prevalence of polypharmacy (five or more) or PIM in dementia, used explicit criteria to identify PIM, and wrote the manuscript in English.The diagnosis of dementia was based on medical records, DSM criteria, ICD code, or other criteria.In addition, the study design of the included articles was observational studies (crosssectional study or cohort study).
Studies were excluded if study subjects had mild cognitive impairment, or if those studies were conference abstracts, reviews, and comments.

Data extraction
Two reviewers (MN Zhao and ZY Chen) independently extracted and verified the data.We extracted information including study characteristics (first author, year of publication, and country), basic information of study subjects (age, sex, and sample size), and study design (setting, prevalence of polypharmacy or PIM, and explicit criteria to evaluate PIM).Any discrepancy between two reviewers was resolved by consultation with a third reviewer (FY, Tian).

Selection of studies
Two reviewers (MN Zhao and ZY Chen) independently screened the titles and abstracts of initially included literature according to the inclusion and exclusion criteria.The full text was further assessed if the eligibility of the study was not clearly determined from the abstract.Any inconsistency in the process of screening was resolved by consulting a third senior investigator (Ting Xu).

Quality assessment
The methodological quality of the cross-sectional study was evaluated using the Agency for Healthcare Research and Quality (AHRQ) (Rostom et al., 2004;Hu et al., 2015).A total of 11 items were listed in AHRQ, including 1) the source of data; 2) eligible criteria for study subjects; 3) time period for included population; 4) whether or not subjects were consecutive; 5) whether the outcome indicators are affected by other factors; 6) any assessments for quality assurance; 7) explanation for excluding any patients from the analysis; 8) measurements taken for controlling confounding factors; 9) description for the handing of missing data; 10) summary for patient response rate and completeness of data collection; 11) clarification of follow-up results.The highest score was 11, while the lowest score was 0. If the score was 8 or above, this study was considered high quality.If the score was 3 or below, this study was considered low quality.If the score was between 3 and 8, this study was considered medium quality.The methodological quality of the cohort study was evaluated by the Newcastle-Ottawa scale (Stang, 2010).If the NOS score ≥8, this study was considered high quality.If the NOS score ≤5, this study was considered low quality.If the NOS score was 6 or 7, this study was considered medium quality (Bedaso and Duko, 2022).

Statistical analysis
We applied STATA, version 16 (Stata Corporation, College Station, Texas, United States) to perform a meta-analysis for polypharmacy and PIM.The pooled prevalence estimate was reported as a proportion with 95% confidence intervals (CI).The I2 statistics was used to assess the magnitude of heterogeneity.When I2 >50%, heterogeneity among studies was considered large and the DerSimonian-Laird random-effect model was applied in analysis.In case of significant heterogeneity, subgroup analysis (e.g., regions, the proportion of females, criteria, and severity of dementia)) was performed to investigate the source of heterogeneity.We also estimated the 95% prediction interval, which further accounts for between-study heterogeneity and evaluates the uncertainty for the effect that would be expected in a new study addressing that same association (Higgins et al., 2009;Migliavaca et al., 2022).Furthermore, a pooled odds ratio was used to analyze the association between PIM and factors when two or more studies reported the same and adjusted odds ratio.Regarding the risk of publication bias, we adopted Egger's and Begg's tests for evaluation.

Quality of the included studies
The results of quality assessment are presented in Supplementary Table S2, 3.For the cross-sectional study, we found that the lowest score was 4, and the highest score was 8. Seven research articles were of high methodological quality (AHRQ score ≥8), and 30 articles were of moderate methodological quality (AHRQ score 3-8).For the cohort studies, 20 research articles were of high methodological quality (NOS score ≥8), four articles were of moderate methodological quality (NOS score 6-7), and one article was of low quality (NOS score ≤5).

FIGURE 2
Prevalence of polypharmacy in older people with dementia across various geographic regions.Note that with <3 studies, the distribution is inestimable and hence not displayed.

Stratified analysis
We applied the stratified analysis for estimating the substantial heterogeneity in pooled prevalence of polypharmacy and PIM.Based on various basic characteristics, such as the proportion of females, year of publication, study design, and severity of dementia, we stratified the studies.
Four studies (Montastruc et al., 2013;Ferreira et al., 2021;Yoon et al., 2022;Zhao et al., 2022) investigated the relationship between polypharmacy (five or more) and the risk of PIMs in older patients with dementia.The pooled estimate was 2.83 (95% CI 1.80-4.44),which indicated that increasing PIM risk was related with polypharmacy.

FIGURE 4
Factors associated with PIM use.HF, heart failure; IHD, ischemic heart disease; COPD, chronic obstructive pulmonary disease; PI, prediction interval.

Publication bias assessment
Egger's and Begg's tests were used to assess the publication bias in the pooled estimate of polypharmacy and PIM.The results for the pooled estimate of polypharmacy were 0.855 (Egger's test) and 0.767 (Begg's test), indicating no publication bias.Regarding PIM, no statistically significant publication bias was observed (Egger test: p = 0.058; Begg's test: p = 0.765).

Discussion
To our knowledge, the current study first comprehensively summarized the pooled prevalence of PIM and polypharmacy in older patients with dementia across different regions and analyzed potential confounding factors associated with PIM use.This review may provide evidence for healthcare decision-makers in avoiding adverse drug use events in elderly with dementia.
For the included studies, the prevalence of PIM varied widely, ranging from 2% to 85.1%.Several reasons could explain the phenomenon well.First, the difference in marketing drugs and medical habits and the gap in healthcare systems in different regions might significantly affect the prevalence of PIM.Based on the pooled results by region, we clearly found a vast difference in PIM prevalence.Renom-Guiteras et al. evaluated the PIM prevalence of eight European countries using the European Union (7)-PIM list and found the PIM prevalence ranging 47%-67.5% (Renom-Guiteras et al., 2018).Zhao et al. also reported differences in PIM prevalence across different cities in China, ranging from 28.48% to 44.79% (Zhao et al., 2022).Second, the screening tools were considered a factor resulting in differences in PIM.To date, many screening tools have been applied to evaluate PIM, such as Beers criteria, STOPP/START criteria, and Holmes criteria (American Geriatrics Society Beers Criteria ® Update Expert Panel, 2019;O'Mahony et al., 2018;Holmes et al., 2008).In our study, a total of 13 criteria were used, of which the most frequently used was Beers criteria, accounting for nearly 50%, followed by STOPP/START criteria.Published PIM lists have important differences in terms of contents and number (e.g., 81 items in STOPP criteria, while 91 in Beers criteria), which might lead to different prevalence of PIM.Target population and clinical practice among various PIM lists might also affect PIM prevalence, such as Holmes criteria mainly focusing on advanced dementia and NORGEP criteria for ambulatory patients (Kaufmann et al., 2014).The stratified analysis based on different criteria in our study also found differences in PIM prevalence.Although the same tool was used, the proportion of patients receiving PIM still varied, which were mainly attributed to how the tools were applied and the edition of the criteria.For instance, several studies just used part of the items of Beers criteria due to the absence of diagnostic information or other laboratory indicators, underestimating PIM to some extent (Chan et al., 2009;Lau et al., 2010;Thorpe et al., 2012;Montastruc et al., 2013).Therefore, before conducting research, researchers must consider how to select appropriate tools and how to apply them based on the collected information and diagnoses.In addition to the impact of region and screening tools on PIM, severity of dementia should be considered.In the analysis, we found that advanced dementia has a lower pooled estimate of PIM than mild-moderate dementia.Despite that patients living with advanced dementia depended completely on others, suffering from a series of distressing symptoms, such as neuropsychiatric symptoms and pain (Moens et al., 2014;Hendriks et al., 2015;Sampson et al., 2018), the emphasis of therapy for those was on ensuring patient comfort and symptom management and reducing polypharmacy (Disalvo et al., 2016).A review by Parsons summarized a viewpoint of physicians about drug use for advanced dementia, recommending discontinuation of anticholinesterase inhibitors, memantine, quetiapine, and simvastatin (Parsons et al., 2010).This may lead to a lower prevalence of PIM in advanced dementia.Despite the vast difference in PIM prevalence among included studies, the pooled estimate of PIM use in our analysis was up to 43%, which was higher than the PIM estimate for older patients in worldwide as given in Tian et al. (2023).Thus, we should pay great attention on the PIM use of older patients with dementia.
The drug management of older patients living with dementia often takes place in the context of additional comorbidities, which result in a large number of prescriptions for patients with dementia (Blass et al., 2008;Callahan and Schubert, 2014;Amy et al., 2018).In the analysis, polypharmacy was found to be prevalent with an estimated overall prevalence of 62%, slightly higher than that found in the study by Janice et al., with an estimate of 59% (Toh et al., 2023).Overall, significant heterogeneity was observed in the prevalence of polypharmacy.The difference may be attributed to several factors, for example, study subjects from different settings, geographical regions, study design, and year of publication.Although heterogeneity did not decrease by subgroups, significant differences were observed between some groups.In our review, we clearly found that the prevalence of polypharmacy in Asia was lowest compared with other regions.This may be due to socioeconomic-related healthcare inequalities between developing and developed countries in the access to healthcare.A report from the WHO declared that developed regions account for 11.6% of the worldwide burden, but account for 90.2% of health expenditure worldwide (Murray and Lopez, 1997).Different settings might affect the prevalence of polypharmacy.In this study, two out of three studies in the Asian region were outpatient studies, with a proportion of only over 10% for polypharmacy.Of note, the same definition of polypharmacy (five or more) shows a difference in measurement, which also might affect the outcome.Lau et al. declared that medications for topical applications, vitamins, and herbal medications were excluded in polypharmacy, with the prevalence of 51.92%, while Bosboom et al. considered all different medications in prescription were being counted, with the prevalence of 92% (Lau et al., 2011;Bosboom et al., 2012).In addition, the time of exposure to the medications has an impact on polypharmacy estimates.Kristensen et al. and Rausch et al. reported the prevalence of polypharmacy at 3 and 6 months as 62.6% and 85.2%, respectively (Kristensen et al., 2018;Rausch and Hoffmann, 2020).Although differences in distribution of regions, measurement, and exposure of time affect the estimate of polypharmacy, polypharmacy still cannot be ignored.
Several studies have been reported regarding potential confounding factors associated with PIM use.The association between PIM and polypharmacy for older patients with dementia was always discussed by researchers.Due to different definitions of polypharmacy and other factors, different studies concluded different results.Ferreira et al. reported polypharmacy was not related to PIM use (Ferreira et al., 2021), while Yoon et al. declared that a strong association between polypharmacy and PIM use was observed (Yoon et al., 2022).In our study, patients with polypharmacy (five or more) were exposed to a higher risk of PIM use, which was consistent with the observation of Tian et al. (2021) (reported older patients).A growing body of studies reported the impact of gender on PIM use, but a consensus has not yet been reached.According to our meta-analysis, no statistically significant difference was observed in women.Of note, we found vascular dementia was more susceptible to PIM use than Alzheimer's disease.This may be related to cardiovascular events being the main cause of vascular dementia (Leys et al., 2005).Cardiovascular diseases and well-recognized high-risk factors of cardiovascular diseases, diabetes and hypertension, were associated with PIM use.This may be due to the use of non-steroidal drugs, regular insulin, and sulfonylureas of the PIM list.Other factors, such as comorbidity, psychosis, depression, and history of cancer, were considered to increase the risk of PIM in the current study.The long-term use of antidepressants, antipsychotics, and opioids in order to control symptoms may explain the phenomenon.Although several factors associated with PIM use were identified in our study, more relevant research was still needed for further validation.
PIM and polypharmacy in older patients with dementia are common.A large number of studies have shown PIM use and polypharmacy were related to hospitalization and death (Gnjidic et al., 2012;Maher et al., 2014).Therefore, it is necessary to optimize drug management for older patients with dementia.Deprescribing is an approach to reduce PIM and polypharmacy (Wu et al., 2021).Due to the use of multiple drugs and memory loss in older patients, drug compliance was relatively poor (Dunbar-Jacob and Mortimer-Stephens, 2001;Smith et al., 2017).For older patients with dementia featuring cognitive impairment and communication disorder, the adherence of drug medication was poorer.Deprescribing could not only reduce the number of medications taken but also increase the drug compliance of elderly patients (Basheti et al., 2016;Jäge et al., 2017).Research has confirmed the advantages of deprescribing in reducing PIM and polypharmacy (Ibrahim et al., 2021).Thus, deprescribing can be applied in the clinics to solve polypharmacy and PIM.
Although the analysis quantitatively summarized the prevalence of PIM and polypharmacy and further explored potential factors with PIM use, providing a reference for the prescription of old patients with dementia, we must acknowledge some limitations of the study.First, those studies included were from all over the world, and other factors, such as culture, education level, geographical location, and social status, would affect the results of polypharmacy and PIM.Second, the included studies show substantial heterogeneity, which may be related to the study subject, sample size, and screening tools used.Third, some studies did not specify the living conditions of the study population and type of dementia, so we cannot conduct a subgroup analysis based on living conditions and type of dementia.Fourth, included studies were limited to English articles, leading to results that could either underestimate or overestimate the prevalence.In addition, only factors using multivariate regression were extracted and several factors in our study were only examined in two studies, which gave biased results.Thus, a further study regarding factors affecting PIM use will be explored.

Conclusion
The analysis revealed that PIM use and polypharmacy were highly prevalent in older patients with dementia.Among different regions, the pooled estimate of PIM use and polypharmacy varied widely.Increasing PIM was related with polypharmacy, women, and vascular dementia.These findings highlight the necessity of some measures taken to improve the prescription quality of older patients with dementia, and they also imply that more caution should be taken when prescribing for women, polypharmacy, and vascular dementia.

FIGURE 1 PRISMA
FIGURE 1 PRISMA flow diagram.PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; PIM, potentially inappropriate medication.

FIGURE 3
FIGURE 3Prevalence of PIM use in older people with dementia across various geographic regions.

TABLE 1
Characteristics of included studies.

TABLE 1 (
Continued) Characteristics of included studies.

TABLE 1 (
Continued) Characteristics of included studies.

TABLE 1 (
Continued) Characteristics of included studies.

TABLE 2
Stratified meta-analysis of the prevalence of polypharmacy and PIM use.

TABLE 2 (
Continued) Stratified meta-analysis of the prevalence of polypharmacy and PIM use.
PIM, potentially inappropriate medication; PI, prediction interval.a One study reported two PIM prevalence using different PIM lists.