Optimizing drug therapy for older adults: shifting away from problematic polypharmacy

ABSTRACT

Introduction

The accelerated discovery and production of pharmaceutical products has resulted in many positive outcomes. However, this progress has also contributed to problematic polypharmacy, one of the rapidly growing threats to public health in this century. Problematic polypharmacy results in adverse patient outcomes and imposes increased strain and financial burden on healthcare systems.

Areas covered

A review was conducted on the current body of evidence concerning factors contributing to and consequences of problematic polypharmacy. Recent trials investigating interventions that target polypharmacy and emerging solutions, including incorporation of artificial intelligence, are also examined in this article.

Expert opinion

To shift away from problematic polypharmacy, a multifaceted interdisciplinary approach is necessary. Any potentially successful strategy must be adapted to suit various healthcare settings and must utilize all available resources, including artificial intelligence.

KEYWORDS:

• Artificial intelligence
• geriatric medicine
• medication-related harm
• multimorbidity
• problematic polypharmacy
• deprescribing

1. Introduction

Physicians have never had more access to pharmaceutical therapeutic agents than they do today. The British National Formulary (BNF), a pharmaceutical reference book containing prescribing information about drugs available in the United Kingdom’s National Health Service (NHS), included approximately 650 medicinal products in the mid-1970s [1]. By 1981, 4500 individual products were included, a figure that has increased manifold in the latest edition of the BNF [2]. This proliferation of pharmaceutical products, referred to as ‘the pharmaceutical revolution,’ has influenced the approach to treating and managing diseases and illness symptoms, reshaping not only the medical framework but also society’s expectations around therapeutics [3]. One significant consequence of this pharmaceutical revolution is problematic polypharmacy, which increases the risk of medication-related harm, especially among vulnerable older adults [4,5]. Although there have been global efforts to reduce problematic polypharmacy, no single reproducible strategy has effectively addressed this challenge. In addition to the marked expansion of pharmaceutical agents over the last 50 years, there has also been an exponential advancement within the medical software sector. Specifically, in recent years, the proliferation of artificial intelligence (AI) driven applications has revolutionized several industries and traditional processes [6]. Artificial intelligence will become increasingly important in healthcare in the coming years, particularly in diagnostic and decision-support applications [7]. In this review article, we will discuss one of the most important consequences of the pharmaceutical revolution – problematic polypharmacy – and suggest how new models of care, supported by innovative AI-based technologies, can help physicians counteract this problem.

2. Multimorbidity and polypharmacy

The expansion of pharmaceutical products has been broadly positive, providing prescribers and patients with an ever broader range of treatment options. Curative treatments are available for many previously lethal illnesses (e.g. antibiotics for many infectious diseases [8], chemotherapy for certain cancers [9]) and disease modifying treatments are available for many chronic, debilitating conditions such as multiple sclerosis [10], rheumatoid arthritis [11] and, most recently, dementia [12]. Many individuals have benefited greatly from increased availability of treatments such as these and are living longer, healthier lives. With life expectancy increasing in most countries since the early 20^th century, there has been a major demographic shift in the world’s population towards old age, an unprecedented modern sociological phenomenon. The World Health Organization estimates that, globally, the number of people over 80 years will treble between 2020 and 2050 reaching an estimated 426 million people [13]. As people age, they are more likely to develop chronic medical conditions, and, when two or more co-morbidities coexist this is referred to as multimorbidity [14,15]. Multimorbidity is associated with worse clinical outcomes, reduced quality of life, and increased healthcare utilization [15–18]. Living with multimorbidity in this era of medication proliferation is associated with increasing levels of polypharmacy. Polypharmacy is commonly defined as the concurrent use five or more different medications [19]. Polypharmacy and multimorbidity are inextricably linked, the relationship being cause-and-effect i.e. an increasing number of diagnosed conditions leads to a proportionate increase in the number of daily medications [20]. Midao et al. analyzed the SHARE (Survey of Health, Ageing and Retirement in Europe) database, which includes detailed information on over 32,000 participants over 65 years in 17 European countries. They found that almost one-third (32.1%) of older adults in Europe take five or more daily medications [21]. Similarly, in the US, Charlesworth et al. examined medication use in adults over 65 years between 1988 and 2010. Over this twenty-two-year period, the prevalence of polypharmacy increased 3-fold, rising to 39% [22]. Trends are similar in Asian countries, where a recent study reported a prevalence of 38.8% and 46.6% in older adults in Taiwan and South Korea, respectively [23]. Other studies have reported varying levels of polypharmacy, ranging from 2.6% to 86.6% [24]. This variation reflects heterogeneity in the definitions of polypharmacy used and the settings in which the studies were conducted. For example, some studies include over-the-counter medications and non-regular medications when assessing whether a patient has polypharmacy.

Polypharmacy is not exclusive to older individuals. Many younger individuals with multimorbidity also experience polypharmacy. However, reduced physiological reserve and age-related changes in pharmacokinetics (i.e. absorption, distribution, metabolism, and excretion) and pharmacodynamics (the predictability of the effect of the drug on the individual patient) predispose older people to medication-related harm [25].

3. Problematic polypharmacy

Polypharmacy is not inherently problematic. Many individuals with multimorbidity are appropriately prescribed five or more medications to manage their chronic health conditions, i.e. appropriate polypharmacy [19]. However, as the quantity of consumed medication increases, there is a proportional increase in risk of patients experiencing medication-related harm in the form of adverse drug reactions (ADRs), adverse drug events (ADEs), drug–drug interactions (DDIs), drug–disease interactions, and drug-geriatric syndrome interactions [4,5]. Problematic polypharmacy encompasses a wide range of phenomena, including inappropriately prescribed medications, inappropriately missed medications, ADRs, avoidable DDIs or drug–disease interactions, inappropriate prescribing cascades, and a failure to align the goals of treatment with the goals of an individual [26–29]. Multiple tools have been developed to help prescribers identify problematic prescribing. Among a large number of explicit screening tools designed to detect problematic polypharmacy are STOPP/START criteria [30], STOPPFrail criteria [31], Beers criteria [32], and EURO-FORTA criteria [33]. Through the use of explicit criteria – clearly defined statements that highlight instances of potentially inappropriate prescribing in particular clinical circumstances – the prescriber is alerted to inappropriate medication use. These tools have been deployed to investigate the prevalence of problematic polypharmacy in multiple settings worldwide, including, acute hospital and community settings and in long-term residential care settings. One study investigating the prevalence of potentially inappropriate medications (PIMs) using STOPP/START version 2 criteria and Beers 2015 criteria identified the presence of one or more PIMs in 66.8% and 54% in community dwelling adults, respectively [34]. Implicit approaches have also been identified as means for assessing medication appropriateness. These include the Medication Appropriateness Index [35], SMART tool [36], and the 10-Step Discontinuation Guide [37]. These tools rely on clinical judgment, expertise, and patient preferences when evaluating medication appropriateness. Due to the complex nature of these interventions, there is limited research investigating these tools in large-scale settings.

4. Factors contributing to problematic polypharmacy

Advancing age and multimorbidity are recognized inter-related factors associated with problematic polypharmacy [20,21]. However, the organization of healthcare delivery is also important. The organization of modern medicine into an individual disease framework giving rise to fragmented healthcare undoubtedly contributes to problematic polypharmacy [38]. Careful consideration of these factors, along with ageing, multimorbidity, and increasingly complex pharmaceutical options, is crucial for a comprehensive understanding of the necessary interventions to address this public health crisis. Individuals living with multimorbidity commonly attend multiple healthcare providers (HCPs) in various different healthcare settings for their medical conditions [38]. This stems from the development of modern medicine around a single disease management framework. Within this framework, HCPs are encouraged to treat each chronic disease in isolation according to evidence-based best practice guidelines. These guidelines are based on evidence from clinical trials, which commonly exclude older multimorbid people [39]. The concern, therefore, is that many treatment guidelines that are applicable to the treatment of individuals with a single chronic disease have the potential to be problematic in multimorbid patients commonly encountered in clinical practice. Application of multiple single-disease treatment guidelines to individual multimorbid patient results in multiple prescriptions, some of which can be ineffective, contradictory, and harmful [40]. This trend toward medication accumulation is further compounded by poorly established communication and coordination structures within fragmented healthcare systems [38].

Evidence-based pharmacotherapy, defined as a decision-making approach guided by appraisal of the best available scientific evidence, aims to ensure that prescribers base their decisions on credible scientific evidence rather than relying on less reliable and potentially biased information such as anecdotal evidence, case reports, or small case series reports [41]. However, the application of all evidence-based guidelines on pharmaceutical products in multimorbid patients can result in evidence-biased pharmacotherapy, i.e. the over-use of evidence-based pharmacotherapy in an undiscerning manner that ultimately amounts to problematic polypharmacy. Factors contributing to evidence-biased pharmacotherapy include faulty methodology [42,43], exclusion of older multimorbid (‘real-world’) patients [39], ambiguity in trial design aspects [43,44], publication bias [44], translation challenges from population-based results to individual patients, and bridging the gap between statistical significance and clinical relevance [45]. Clinical trials often employ specific inclusion and exclusion criteria, leading to recruitment of participants who may not represent the general population [46]. Additionally, clinical trials are frequently designed by pharmaceutical companies with a vested interest in positive results. Sometimes, negative clinical trial results may not be subsequently put forward for publication, a problem that is compounded by publication bias, wherein trials with favorable outcomes are more likely to be published [44]. Although considerable efforts have been made in recent years to ensure research integrity and to increase the reporting of ‘negative’ trials results with balanced interpretation of trial data, considerable work is still required in this domain [45,47].

The International Union of Basic and Clinical Pharmacology (IUPHAR) geriatric committee has recognized the significant gap in the inclusion of older adults in research literature and has published strategies to improve medication safety and efficacy for this cohort [48]. The committee recommends improvements in clinical trial design, emphasizing the inclusion of adults with frailty and the measurement of geriatric syndromes, such as falls, frailty, and cognition, both at baseline and as outcomes. This approach is particularly important when older adults are the intended recipients of the medication. The committee advocates for a pragmatic approach to trial design to facilitate the inclusion of frail older adults, suggesting provisions such as home visits and video calls with caregivers. By adopting an improved, collaborative, inclusive, and pragmatic trial design, older people can be better represented in research. This will ultimately lead to enhanced medication safety and appropriateness for older adults.

5. Consequences of problematic polypharmacy

Problematic polypharmacy is associated with a range of adverse consequences for individual patients, including increased risk of ADRs, increased hospitalization, reduced health-related quality of life, and ultimately, increased mortality [26–28,49–51]. A UK prospective, observational study reviewed all patients admitted via the medical assessment unit over a one-month period [50]. The authors reported that ADRs were the main reason for or a contributing factor in 18.4% of hospital admissions and that those admitted with an ADR were older, had more co-morbidities, and were more likely to have polypharmacy than patients admitted for reasons other than ADRs [50].

Several studies have demonstrated a significant association between polypharmacy and mortality. A systematic review and meta-analysis by Leelakanok et al. identified a dose–response relationship, with increasing numbers of medications being significantly associated with increased mortality (p < 0.05) [51]. Older women are intrinsically more vulnerable to experiencing these negative consequences, due to intrinsic differences in drug pharmacodynamics and pharmacokinetics compared to older men [52,53]. In addition, beyond the age of 75 years, women outnumber men by approximately 2:1 in most developed countries, a survival difference that becomes more accentuated the older the age cohort [54].

Polypharmacy and resultant medication-related harm place a considerable economic burden on healthcare systems through increased drug expenditure and expenses related to treating medication-induced harm. Research conducted by the Intercontinental Medical Statistics (IMS) Institute for Healthcare Informatics, entitled ‘Advancing the Responsible Use of Medicines: Applying Levers for Change’ reported that approximately 4% of the world’s total avoidable healthcare cost can be attributed to mismanaged or suboptimal prescribing [55]. A Canadian study by Morgan et al. reported that the use of PIMs can increase the annual per-patient costs by approximately $2000, resulting in an estimated total annual cost of $419 million to the country [56]. Similarly, a Japanese study by Akazawa et al. reported a significant financial impact from PIM prescribing, with older adults prescribed PIMs incurring 33% average higher medical cost than those not exposed to PIMs [57]. In the UK, it is estimated that the NHS spends approximately £2.21 billion annually as a result of ADR-related hospital admissions, most of which occur in multimorbid older people on long-term polypharmacy [50].

6. WHO challenge

The growing body of evidence highlighting iatrogenic harm caused by problematic polypharmacy prompted the launch of the third World Health Organisation (WHO) patient safety challenge ‘Medication without Harm’ in 2017 [58]. Through this initiative, the WHO strove to raise awareness about the importance of medication safety and to seek new and specific commitments from global leaders in healthcare to make policy changes to support safer medication practices. The specified target was to reduce severe and avoidable harm from medication by 50% within five years. The WHO challenge aimed to improve medication safety at each stage of the medication process, from prescribing, to dispensing, to administering, to monitoring, and utilization of drug therapy. It specifically targeted three contexts where patients face a particularly high risk of experiencing medication-related harm. These contexts were to be the basis for change implementation and included high-risk situations, polypharmacy, and transitions of care. High-risk situations encompass scenarios which pose a significant risk of medication-related harm to patients, such as the use of higher-risk medications like insulin and warfarin and interruptions during critical processes such as prescribing and administration of medication. Transitions of care from community to hospital and later from hospital to home are commonly associated with unintentional medication discrepancies that can be potentially harmful. Older individuals with multimorbidity frequently encounter all three contexts (high-risk situations, polypharmacy, and transitions of care) simultaneously, compounding the risk of adverse drug events in this cohort. The challenges of optimizing medications in the heterogenous group of multimorbid older adults are evidenced by the fact that over five years since this ambitious goal was set, there is no published evidence to indicate that the initiative has significantly reduced medication-related harm, as expressly intended by the WHO. To emphasize the persistent necessity of achieving this goal, the WHO Safety Action Plan 2021–2030 prioritizes the Medication without Harm Challenge in one of its key strategies (strategy 3.2) [59]. This strategy outlines key actions for governments, healthcare, stakeholders, and the WHO secretariat, including establishing expert groups and leadership teams, appointing medication officers/teams in healthcare, promoting patient involvement and medication literacy, recognizing and reporting of ADRs, and monitoring of progress. A strategic and coordinated approach is recommended to work toward this goal.

7. Evidence-based interventions to reduce harm associated with problematic polypharmacy

Since the WHO launched this challenge, several large-scale multicentre trials have investigated novel interventions designed to reduce problematic polypharmacy. A cluster randomized clinical trial in Canada, the MedSafer Study, utilized an electronic deprescribing support tool which created personalized deprescribing reports in older adults with polypharmacy [60]. MedSafer enrolled 5698 hospitalized participants who were randomized to standard care or personalized deprescribing reports. This study’s aim was to evaluate the effect of this support tool on adverse drug events after hospital discharge. For the primary outcome in MedSafer, ADEs, there was no significant difference between the treatment groups. However, use of the deprescribing support tool successfully identified opportunities for deprescribing, although uptake of these recommendations by attending clinicians was variable. In the case of certain medications with high side effect profiles in older patients, substantial levels of deprescribing took place in the intervention cohort i.e. benzodiazepines and sedative hypnotics, gabapentinoids, and trazadone were deprescribed in 40%, 35%, and 30% of users, respectively.

Low implementation of prescribing advice points was also a feature of the SENATOR trial, a large multicentre, randomized control trial, which enrolled 1537 hospitalized patients [61]. Patients were randomized to usual pharmaceutical care or to the intervention arm, which involved provision of individualized software generated reports with prescribing advice points to attending doctors. The prescribing advice was largely based on STOPP/START version 2 criteria. There was no significant difference in the incidence of ADRs as the primary endpoint between the groups, not surprising, given that only 15% of the advice points were implemented in the intervention patient cohort. An accompanying qualitative study [62] investigating factors that influenced prescribing advice implementation identified a number of key factors that mitigate against prescribing advice implementation, including low clinical relevance of some recommendations in the acute setting, time constraints in a busy hospital setting, prescribers level of experience, and patient-specific issues. Failure of the MedSafer and SENATOR trials to show significant reductions in the incidence of ADRs in the intervention groups was fundamentally linked with suboptimal implementation of prescribing advice.

Challenges with reducing PIMs were also a feature in the SPPiRE trial [63]. This was a community-based cluster randomized controlled trial that enrolled 404 participants. The intervention was delivered by general practitioners (GPs) who underwent pre-trial training on polypharmacy management in older people. PIMs were identified by GPs based on learning through training videos. Medication changes were implemented by GPs using clinical judgment. There was no change in the number of PIMs prescribed or other outcome measures in the intervention patient cohort compared to the control cohort.

These trials underscore the complexity of deprescribing PIMs both in the acute hospital setting and in the community. Furthermore, they highlight the challenges of implementing medication changes when recommendations are generated by an external source or by primary care physicians who know their multimorbid older patients in greater detail.

A cumulative update, i.e. a review of systematic reviews evaluating interventions designed to improve polypharmacy management was recently published by Keller et al. in which fourteen systematic reviews, accounting for 179 unique published studies were evaluated [64]. Their results highlighted the ongoing lack of sufficient high-quality evidence to endorse the widespread implementation of interventions targeting polypharmacy. The authors noted that detailed comprehensive medication reviews (involving prescription review, medication adherence review, and a face-to-face review of medications and conditions with the patient) were the most effective interventions and the only intervention associated with a reduction in unplanned hospital admissions.

As highlighted, the absence of high-quality evidence around interventions to address problematic polypharmacy does not stem from a lack of substantial investment in this area. Furthermore, the inherent heterogeneity of multimorbid older adults with complex medical conditions and associated problematic polypharmacy complicates the response patterns of prescribers to interventions. The results from large, multi-center, trials such as Senator and MedSafer, have shown that simply applying explicit PIM criteria such as STOPP/START criteria or Beers criteria to medication lists may not result in a meaningful improvement in clinically important patient outcomes. Some smaller, single-center studies have reported benefits in patient outcomes. O’Connor et al. demonstrated that the application of STOPP/START criteria to medication lists of hospitalized older adults resulted in significantly fewer ADRs compared with usual pharmaceutical care [65]. Garcia et al. highlighted that patients who received a deprescribing intervention were less likely to experience falls, delirium, and hospital admissions [66]. Similarly, Frankenthal et al. reported a reduction in falls in patients who received a deprescribing intervention [67]. The variability of results likely relates to factors beyond simply applying explicit criteria. For example, in smaller studies implementation rates tend to be higher, for example, O’Connor et al. reported a implantation rate of 83.4% [65]. In the study by Garcia et al., the patient’s own physicians were trained to identify PIMs using explicit criteria. These physicians, who had established relationships with patients, likely adopted a nuanced approach to deprescribing informed by their own understanding of the individual patient histories. Moreover, their authority to deprescribe eliminated challenges with implementation [66]. This emphasizes the necessity for a multifaceted, personalized approach to successfully improve problematic polypharmacy in this high-risk cohort of patients. The Geriatrics 5 Ms framework has been proposed as a method of approaching medication optimization in older adults [68]. The 5 Ms brings focus to broader aspects of older person care, including Mobility, Mind, Multicomplexitity, What Matters Most, in addition to Medications. The goal is to provide an Age-Friendly Health System where these elements of high-quality care are reliably provided across healthcare settings. While elegant and holistic, it is a framework for high-quality care rather than a specific tool or procedure for addressing complex polypharmacy.

8. Polypharmacy stewardship

A new concept referred to as ‘polypharmacy stewardship’ addresses many of the shortcomings of previously described interventions and has been proposed as a model of care to reduce the risks associated with problematic polypharmacy in older people [69]. Polypharmacy stewardship is defined as a coordinated intervention designed to assess, monitor, improve, and measure the pharmacotherapeutic treatment of multimorbidity, taking into account potentially inappropriate medications, potential prescribing omissions, potentially or actually detrimental drug – drug and drug – disease interactions, and prescribing cascades, with the aim of aligning treatment regimens with the overall condition, prognosis, and preferences of the individual patient. The concept of this stewardship model is adapted from the principles of antimicrobial stewardship, which has been shown to improve both patient and economic outcomes [70–73]. A polypharmacy stewardship program includes the following core components: optimization of the technical components of prescribing, personalized prescribing, and deprescribing in view of realistic treatments goals, avoidance of medication-related harm, patient-centered care with provision of supports and follow-up as necessary and collaboration with key stakeholders including community-based health care providers (Figure 1). Several of these core components align closely with the ten recommendations published by the International Group for Reducing Inappropriate Medication Use & Polypharmacy (IGRIMUP), such as recommendation four, ‘employ mixed implicit and explicit approaches to polypharmacy,’ and recommendation ten, ‘decisions in older complex patients should routinely consider expected survival and quality of life, giving the highest priority to patient/family preferences’ [74]. There is no single evidence-based preferred operational model for a polypharmacy stewardship program. Nevertheless, as healthcare providers strive to implement the key actions outlined in the WHO Safety Action Plan 2021–2030 [59], a polypharmacy stewardship program serves as an effective strategy to meet these targets. Initial steps to integrate a stewardship program into practice include securing leadership support, establishing accountability and reporting structures, appointing physicians and pharmacists with appropriate expertise, and monitoring outcomes. Successful implementation of polypharmacy stewardship across a diverse range of healthcare settings is theoretically feasible and requires adaptation of the core components while considering the resources available and the financial viability of the program.

Figure 1. Process model for polypharmacy stewardship. PIMs =Potentially inappropriate medications, PPOs= potential prescribing omissions, ADRs= adverse drug reactions – adapted from [69] under creative commons CC-BY-NC-ND license.

9. Artificial intelligence (AI)

One rapidly evolving resource that is anticipated to play an increasing role in polypharmacy management is AI [7]. To date, AI-based advances in healthcare have been most prominent in radiology and surgery [75–78], where the application of AI-based techniques has shown improved patient outcomes compared to conventional clinical practice. The Radiology Health AI register provides an overview of AI software available in Europe for clinical radiology practice. Over 200 products are currently listed that can enhance diagnostic precision and identification of abnormalities on imaging [79]. AI in surgery is currently transforming traditional surgical treatment approaches. Robotic surgical systems are being increasingly utilized in specialties such as urology and gynecology where they offer increased precision of surgical intervention, safer procedures and improved patient outcomes compared to traditional surgical treatment [76–78,80].

The role of AI in supporting physicians to manage polypharmacy has been evaluated more recently. In a 2023 single-center study, Akyon et al. described the development of a rule-based auxiliary tool and AI supported web application to aid physicians’ management of polypharmacy [81]. This tool was created to incorporate patients’ age, medications, and comorbidities and applies prescribing criteria and guidelines. Drugs were screened according to six PIM criteria and classified as ‘usable,’ ‘risky’ and ‘no warning found.’ Medications were also screened for and classified according to drug–drug interaction and drug–disease interactions. Medications were evaluated in 296 older adults. The comparison between physicians and the AI tool revealed a substantial difference in the time taken to detect PIMs, PPOs, and DDIs, with physicians averaging 38 min while the AI tool took only 34 s.

Dil Nahlieli et al. validated an Artificial Pharmacology Intelligence (API) system [82]. The API system integrated data from patients’ health records and other sources, incorporating medications, medical history, and personal parameters. This enabled the API system to generate prioritized personalized assessments of drug-related risk in accordance with up-to-date medical literature and guidelines. API system significantly outperformed clinical pharmacist reviews (p = 0.0001) in identifying medication-related problems. Both these studies underscore the potential of AI as an aid to healthcare professionals, particularly in synthesizing extensive amounts of data efficiently.

TIME (Turkish inappropriate medication use in the elderly) criteria were developed by a group of pharmaceutical experts in Turkey and include recommendations for both prescribing and deprescribing medications in older people [83]. Recognizing the challenge of integrating comprehensive criteria into busy clinical settings, the group designed and developed a smartphone application of TIME enabling clinicians to efficiently deploy the tool in clinical practice.

Other studies have utilized AI technology to improve dosing in patients using warfarin, as well as predicting DDIs in patients through a framework which integrated phenotypic, chemical, therapeutic, and genomic properties [84,85].

These initial research findings indicate that AI can support prescribers by synthesizing large volumes of drug safety and interaction data on a practical and accessible platform. However, it is crucial to also consider the challenges associated with AI. Key challenges include understanding the complexity of AI algorithms, overcoming technical issues, mitigating bias, ensuring data security, and avoiding over-dependence on technology. Notwithstanding these challenges, given the scale of the task of improving medication safety, leveraging advanced technologies to tackle complex polypharmacy seems unavoidable. With appropriate safeguards and governance, AI has the potential to enhance decision -making and efficiencies in the care of patients with complex polypharmacy.

10. Conclusion

Problematic polypharmacy is a growing concern with increasing prevalence internationally despite a global effort to reduce this problem for over five years. Failure to successfully reduce problematic polypharmacy in multimorbid older adults stems largely from the challenges of tackling a multi-factorial problem in a rapidly growing heterogenous population. Successful interventions will need to be multi-faceted, involve face-to-face contact with multimorbid older patients, and will likely benefit from incorporating AI technologies. We contend that such an intervention can be developed and deployed successfully, which will improve outcomes for millions of older adults living with problematic polypharmacy worldwide.

11. Expert opinion

Problematic polypharmacy is a major healthcare and patient safety concern worldwide, stemming from multiple factors including ageing populations, multimorbidity, fragmented healthcare systems, over-emphasis on evidence-based pharmacotherapy, and the organization of healthcare around an individual disease treatment framework. The multi-layered, multifactorial nature of polypharmacy in late life underscores its complexity, making it challenging to manage. Nonetheless, it is imperative that progress is made in reducing problematic polypharmacy due to its wide-ranging negative consequences for patients, healthcare systems and macro-economics. Therefore, any intervention targeting problematic polypharmacy should aim to improve outcomes across these domains. Despite numerous randomized controlled trials seeking to address problematic polypharmacy, successes have been limited in terms of impact on clinically relevant endpoints. That is, while many of these clinical trials, such as SENATOR [61] and MEDSAFER [60] have demonstrated methods that identify PIMs and PPOs to prescribers, implementation of advice points has been poor, and they have generally failed to demonstrate clear improvements in patient outcomes such as hospital readmission, medication-induced impairment of quality of life and mortality. A significant issue lies in the complexity of the polypharmacy-attenuating interventions in question and the financial resources needed to support them. Unlike many other areas of pharmacotherapy where pharmaceutical companies promote and fund pharmacotherapy research, interventions focusing on deprescribing will struggle to attract such support. While this helps maintain research integrity, it poses a challenge to researchers to successfully undertake polypharmacy-improving research through federal and state government and philanthropic funding sources.

Another obstacle with determining successful interventions to target problematic polypharmacy is the complexity of patients living with multimorbidity-induced polypharmacy and the complex multitude of factors that contribute to problematic polypharmacy. To address this challenge, we propose polypharmacy stewardship as a novel collaborative, multifaceted, intervention approach. The framework for polypharmacy stewardship is modeled on the anti-microbial stewardship program which has been applied successfully in many countries to improve patient and economic outcomes related to antimicrobial pharmacotherapy. A key component of polypharmacy stewardship involves optimizing the technical aspects of prescribing, encompassing optimal dosing, formulation, dispensing, as well as identification of PIMs and PPOs, DDIs, and prescribing cascades. It is within this domain that artificial intelligence (AI) will likely demonstrate its practical value, particularly in reducing the time taken to identify prescribing errors, potentially inappropriate prescribing (IP), and potential medications for deprescribing. AI technologies can be applied to operationalize and automate IP criteria such as STOPP/START criteria [30] and Beers criteria [32] enabling prescribers to streamline the identification of PIMs and PPOs. Integration and utilization of such technologies would then afford physicians more time to focus on more nuanced aspects of polypharmacy, such as establishing patient preferences, involving patients in the decision-making process to a greater extent, and providing follow-up. While initial research has been undertaken, further research is required to fully explore the potential benefits of AI within a polypharmacy stewardship program. Without leveraging the clinical potential of AI, the feasibility of funding interventions that require significant manpower will remain a challenge to researchers and healthcare systems alike. We expect that the integration of AI into polypharmacy stewardship programs will help to optimize resource allocation and improve patient and economic outcomes.

Looking ahead to the next 10–20 years, polypharmacy is expected to become more prevalent in tandem with ageing demographic changes which will inevitably generate higher levels of multimorbidity and its by-product, polypharmacy. A more focused, better structured and funded approach to polypharmacy research is needed to elucidate the theoretical value of multifaceted interventions like polypharmacy stewardship, augmented by innovative AI-based technologies. If such approaches are proven to be successful, the patient safety and economic dividends could be very substantial indeed.

Article highlights

• Over the last 50 years, there has been an unprecedented proliferation of medications available. This has influenced the approach to treating and managing diseases and illnesses and is one of the drivers of polypharmacy in modern medicine.

• Problematic polypharmacy encompasses a range of phenomena, including inappropriately prescribed medications, avoidable drug–drug interactions or drug–disease interactions, and inappropriate prescribing cascades.

• Problematic polypharmacy is associated with a range of adverse patient outcomes, including ADRs, increased hospitalization, reduced health-related quality of life, and ultimately, increased mortality.

• Despite significant investment, no trial has yielded an intervention that is reproducible and effective in improving outcomes for older adults with problematic polypharmacy. This underscores the complexity of deprescribing in a heterogenous, complex, and multimorbid population.

• Further clinical trials are required to examine novel interventions that optimize drug therapy. Successful interventions will likely be multifaceted, interdisciplinary, patient centered and will utilize new technological resources such as AI.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.