This preprint asks a deceptively simple question with system-level consequences: what actually works in deprescribing when you move beyond slogans and look across settings, professions, and intervention designs (Manley et al., 2025). Deprescribing is framed here not as a niche geriatric practice, but as a public health strategy for reducing medication-related harm, waste, and avoidable environmental burden that stems from overuse and inappropriate prescribing (Manley et al., 2025). The authors start from the premise that rising medicine consumption and high rates of potentially inappropriate prescribing, especially among older adults and residents in long-term care, have created a layered problem. It is clinical because polypharmacy increases adverse drug reaction risk; it is economic because unnecessary prescribing and downstream complications increase spending; and it is environmental because unused medicines and pharmaceutical production and disposal contribute to pollution and greenhouse gas emissions (Manley et al., 2025). Their central purpose is to map the intervention landscape and provide an initial effectiveness signal across that landscape, so policymakers and implementers can stop improvising in the dark and start scaling what looks reliably functional (Manley et al., 2025).
Methodologically, the work is a scoping review designed to capture breadth rather than produce a single pooled effect size. The authors follow established scoping review methodology and reporting guidance, search PubMed and Web of Science from 2010 onward, and include interventional studies that explicitly aim to reduce or stop medication use rather than merely “improve prescribing quality” in a general sense (Manley et al., 2025). They deliberately avoid grey literature and exclude protocols and qualitative-only studies, aiming for evaluated interventions with reported outcomes that allow at least directional judgments about impact (Manley et al., 2025). A practical, modern detail is that they apply GRADE criteria to assess evidence quality, assisted by ChatGPT (version 4), with additional author checking for consistency (Manley et al., 2025). That matters because their headline claims about “what works” are repeatedly anchored in whether the higher-quality segment of the evidence aligns with the same design features.
The dataset they assemble is large enough to reveal patterns. From an initial pool of 596 records, they include 179 studies after screening and eligibility assessment (Manley et al., 2025). The literature is recent and accelerating: most included studies are published after 2017, and a substantial share after 2020, suggesting a field shifting from conceptual advocacy to implementation experimentation (Manley et al., 2025). Geographically, publications concentrate in high-income countries, with the United States and Canada prominent, and recognizable intervention “clusters” appearing repeatedly, which implies that a few scalable models have been tested, adapted, and re-tested in multiple contexts (Manley et al., 2025). Study designs are mixed: randomized controlled trials are common, but so are pre–post and feasibility designs, which expands the map but also explains why the authors treat causal certainty cautiously (Manley et al., 2025).
Where the paper becomes especially useful for decision-makers is how it organizes interventions by setting, target, and components. Most interventions occur in community or outpatient environments, with substantial representation in inpatient care and nursing homes, and only a small number in emergency departments (Manley et al., 2025). Populations are overwhelmingly adult, and heavily skewed toward older adults, with additional targeting of polypharmacy but variable thresholds for what counts as polypharmacy across studies (Manley et al., 2025). That variability is not an academic nuisance; it is a scaling problem. If one system defines polypharmacy as five or more drugs and another uses fifteen, “successful deprescribing” will represent very different baseline risk and clinical complexity, making replication and benchmarking harder (Manley et al., 2025). The authors also identify disease- or context-specific targeting (e.g., dementia, chronic disease, psychiatric disorders), but the main gravitational center remains older adults and multi-medication burden (Manley et al., 2025).
The paper’s clearest design signal is about who the intervention aims to move. The majority of interventions are provider-focused, fewer mix provider and patient components, and a minority are primarily patient-focused (Manley et al., 2025). This matters because deprescribing is a behavioral and organizational act as much as a pharmacological one: someone must decide, someone must agree, someone must implement tapering and monitoring, and someone must absorb the perceived risk of stopping what was once started. Provider-focused approaches typically include education, medication review protocols, and decision-support tools. Mixed interventions add patient-facing educational or empowerment components. Patient-focused interventions often rely on structured educational materials designed to shift patient preferences and prompt conversations with clinicians (Manley et al., 2025). The paper highlights the EMPOWER brochure as a recurring patient empowerment instrument, frequently used for benzodiazepine deprescribing, and sometimes adapted to other classes (Manley et al., 2025). The underlying message is blunt: if you only change knowledge but not interaction dynamics, deprescribing stalls; if you change interaction dynamics with good tools, it moves.
Professional configuration emerges as another strong pattern. Nearly half the studies are multidisciplinary, and pharmacists appear as pivotal actors within multidisciplinary teams and also as single-discipline intervention leaders in many studies (Manley et al., 2025). This is not presented as “pharmacists are better,” but as an implementation observation: pharmacists are structurally positioned to run medication reconciliation, identify deprescribing targets using explicit criteria, and operationalize tapering guidance, especially when embedded into collaborative workflows with physicians and nursing staff (Manley et al., 2025). The paper also catalogs the medical specialties that commonly participate, which implicitly signals where deprescribing has become normalized as part of routine practice rather than perceived as an exceptional deviation (Manley et al., 2025).
Target medication strategies split into broad versus class-specific approaches. About half the studies do not focus on a single drug class, instead using criteria such as Beers or STOPP to identify potentially inappropriate medications (Manley et al., 2025). The other half often targets “usual suspects” where harm–benefit ratios are frequently unfavorable or dependence and long-term risks are salient, including benzodiazepines, anticholinergic/sedative drugs, and proton pump inhibitors (Manley et al., 2025). Crucially, the operational backbone of targeting is frequently guideline-based: roughly half of studies use explicit guidelines, with Beers and STOPP appearing most often, and sometimes integrated into electronic medical records (Manley et al., 2025). This is an important systems insight: scaling deprescribing is easier when the “what should we stop” question is not reinvented each time by individual clinicians, but supported by standardized criteria and automated prompts.
When the authors break down intervention content, five components dominate: provider education, patient education, medication review, clinical decision support systems, and (less commonly) behavioral therapy or medication substitution as a bridge during tapering (Manley et al., 2025). Medication review is the most pervasive element, present in a large majority of studies, and delivered via multiple modalities: face-to-face, phone, EMR-based, home visits, or hybrid approaches (Manley et al., 2025). Clinical decision support tools are particularly notable because they represent a shift from artisanal deprescribing to semi-automated identification and individualized recommendations, exemplified by systems that generate patient-specific reports, flag high-risk prescriptions, and provide tapering instructions (Manley et al., 2025). The practical implication is not “buy software,” but “make deprescribing easier than doing nothing,” because default inertia is powerful in prescribing behavior.
Outcomes tell a story about what the field currently values and what it avoids. Most studies measure medication-level outcomes such as number of drugs discontinued or reduction rates, and many stop there (Manley et al., 2025). Patient-level outcomes are also common, including adverse drug events, mortality, hospitalizations, emergency visits, and quality of life, but the paper reports that negative patient-level clinical effects are not observed in the included studies, with only limited transient issues resolved by reintroducing medications in a small number of cases (Manley et al., 2025). Economic outcomes appear in a smaller subset; cost savings are reported in about a third of those, while others show no change or mixed results (Manley et al., 2025). The most striking gap is environmental outcomes: despite the paper’s strong framing that medication overuse has a substantial environmental footprint, none of the included intervention evaluations quantify environmental benefits such as reduced pharmaceutical waste or reduced carbon footprint (Manley et al., 2025). In policy terms, this is a measurement failure that blocks cross-sector budgeting. If deprescribing is supposed to support planetary health, the field needs indicators that allow health systems to claim and track those gains.
The authors attempt to move beyond “everything works” optimism by anchoring interpretation to evidence quality. Using GRADE, they report that only a minority of studies sit in moderate to high quality, with many low or very low due to design limitations typical in implementation work (Manley et al., 2025). Yet within the moderate-to-high group, about three-quarters show significant impact, and several design features appear repeatedly in successful interventions: stepwise approaches, structured medication review protocols, and guideline-based decision-support tools, including STOPP/Beers-linked systems (Manley et al., 2025). Multidisciplinary interventions also show a high share of significant effects, with pharmacists frequently involved (Manley et al., 2025). Mixed provider–patient interventions show more variable success; when they work, they tend to include genuine patient empowerment tools rather than generic information sheets (Manley et al., 2025). The practical reading is that deprescribing is not fragile, but it is design-sensitive: interventions succeed when they reduce cognitive load, clarify targets, assign roles, and create a repeatable workflow.
The discussion section expands the “purpose” from clinical optimization to system strategy. The authors argue that the evidence supports deprescribing as safe and effective, and that the real challenge is scaling it through policy design that is context-sensitive rather than one-size-fits-all (Manley et al., 2025). They place particular emphasis on embedding deprescribing into coherent national or regional policy, aligning incentives, building coordination infrastructure, and sustaining behavior change through monitoring and feedback loops (Manley et al., 2025). This framing is consequential: deprescribing becomes less like a project and more like a capability. In that sense, the paper’s underlying argument is that effective deprescribing resembles effective quality improvement: it needs standardized tools, multidisciplinary governance, and patient-involved routines, not heroic individual clinicians.
Their stated future directions also reveal where the next research frontier should move if the field is serious about public health impact. The literature is heavily skewed toward older adults, leaving working-age populations and earlier prevention of long-term polypharmacy under-explored (Manley et al., 2025). Equity and heterogeneity are weakly addressed: few studies test whether interventions work similarly across socioeconomic, ethnic, linguistic, or geographic groups, which is a major problem if scaling is the goal (Manley et al., 2025). Follow-up time horizons are often short, limiting inference about sustainability of medication reductions and longer-term utilization outcomes (Manley et al., 2025). And the environmental dimension remains unevaluated, which is the most ironic gap given how central it is to the paper’s motivation (Manley et al., 2025). The authors are explicit about the limitations inherent in scoping reviews and in heterogeneous intervention evidence, including publication bias toward positive findings and the inability to meta-analyze across incomparable designs (Manley et al., 2025). The net effect is that the paper delivers a strong map and reasonable directional guidance, while warning against overconfident ranking of intervention types.
If you want one sentence that captures the paper’s purpose without wasting syllables, it is this: deprescribing works often enough, safely enough, and across enough settings that the question is no longer whether deprescribing should be pursued, but how to design and institutionalize it at scale using multidisciplinary collaboration, patient engagement, and guideline-based decision tools, while fixing the glaring measurement gaps around long-term and environmental outcomes (Manley et al., 2025). That is a public health agenda disguised as a scoping review, which is probably why it lands.
Mini dictionary of key concepts
Deprescribing is the planned, supervised reduction or cessation of one or more medicines to improve patient outcomes by reassessing benefit–risk and identifying low-value or inappropriate prescriptions, then executing a discontinuation plan rather than leaving medication accumulation to inertia (Manley et al., 2025). In this paper, deprescribing is positioned as both a patient safety intervention and a system-level strategy with economic and environmental relevance (Manley et al., 2025).
Polypharmacy refers to the use of multiple medications, but the review shows that the operational definition varies widely across studies, commonly using thresholds ranging from five to fifteen drugs (Manley et al., 2025). This variability matters because it changes who qualifies for interventions, how “success” is counted, and how comparable findings are across settings.
Potentially inappropriate medications are drugs for which harms likely outweigh benefits in certain populations or contexts, often identified through explicit criteria such as Beers or STOPP (Manley et al., 2025). In the reviewed interventions, these criteria function as scalable targeting rules that reduce subjectivity and help translate deprescribing into routine practice (Manley et al., 2025).
Medication review is a structured assessment of a patient’s medication regimen, often used as the first step in stepwise deprescribing processes, and performed by pharmacists, physicians, or multidisciplinary teams across inpatient, outpatient, nursing home, and hybrid settings using interviews, EMR review, or both (Manley et al., 2025). The review suggests medication review is the dominant operational mechanism through which deprescribing decisions become actionable (Manley et al., 2025).
Clinical decision support systems are digital tools that identify deprescribing targets and generate patient-specific recommendations or alerts, sometimes embedding guideline logic into electronic records and producing tapering instructions to reduce workload and improve consistency (Manley et al., 2025). In the higher-quality evidence subset, guideline-based decision support is repeatedly associated with successful medication reduction outcomes (Manley et al., 2025).
Patient empowerment tools are structured educational resources designed to shift patient beliefs and prompt deprescribing conversations, with the EMPOWER brochure repeatedly used as a standardized mechanism, especially for benzodiazepine deprescribing and sometimes for other classes (Manley et al., 2025). The review’s pattern is that patient involvement is most effective when it is instrumented with concrete tools rather than generic advice (Manley et al., 2025).
GRADE is an evidence quality appraisal framework used here to classify the strength of included studies, with many studies downgraded due to methodological limitations and heterogeneity, and a smaller subset achieving moderate-to-high ratings (Manley et al., 2025). The authors use this to avoid treating all positive results as equally credible, and to identify which intervention features appear robust within stronger study designs (Manley et al., 2025).
References: Manley, E., Or, Z., & Brunn, M. (2025). What may work in deprescribing? A scoping review of intervention types, targets and outcomes (medRxiv preprint). https://doi.org/10.1101/2025.07.30.25328075