Screening an elderly patient's medication list against the AGS Beers Criteria for potentially inappropriate medications
Older adults are prescribed medications at a higher rate than any other age group, and the physiologic changes of aging — reduced renal clearance, altered volume of distribution, increased CNS sensitivity — mean drugs that are safe in younger adults can become inappropriate in later life. The American Geriatrics Society (AGS) Beers Criteria is the most widely used explicit tool for identifying these potentially inappropriate medications (PIMs) before harm occurs.
Geriatrician Mark H. Beers first published explicit criteria for inappropriate medication use in nursing home residents in 1991. The list was revised in 1997 and 2003, then adopted, expanded, and taken over by the American Geriatrics Society in 2011. Since then AGS has published updates in 2012, 2015, 2019, and most recently 2023, each produced by an interdisciplinary expert panel using a modified Delphi consensus process grounded in systematic literature review.
The 2023 update reviewed evidence published since 2015, added and removed several drugs and classes, and reorganized content into five clearly separated criteria tables — the same five categories this simulator screens against. The criteria are endorsed by CMS quality measures, incorporated into most EHR clinical decision support, and used by pharmacists, prescribers, and researchers worldwide as a shorthand for prescribing quality in older adults.
The Beers Criteria are explicitly NOT meant to be applied as a rigid checklist. AGS states the criteria are intended to inform, not dictate, clinical judgment — a flagged medication may still be the right choice for an individual patient after weighing risks, benefits, and preferences.
Several physiologic shifts converge to make many ordinary medications riskier in older adults:
• Pharmacokinetics: reduced hepatic blood flow and phase I metabolism slow drug clearance; reduced glomerular filtration rate (GFR declines ~1%/year after age 40) prolongs half-life of renally-cleared drugs like gabapentin and many antibiotics • Pharmacodynamics: the aging brain is more sensitive to anticholinergic and sedating drugs — the same dose of diazepam produces deeper, longer sedation in an 80-year-old than a 40-year-old • Body composition: decreased lean mass and total body water increase volume of distribution for lipophilic drugs (benzodiazepines), prolonging their effect • Homeostatic reserve: reduced baroreceptor sensitivity and postural blood pressure control means antihypertensives and psychoactive drugs are more likely to cause orthostatic hypotension and falls
On top of the biology, older adults also take more drugs: the average Medicare beneficiary fills prescriptions from multiple prescribers who may not see each other’s medication lists, creating fertile ground for both PIMs and dangerous interactions.
Polypharmacy — conventionally defined as five or more concurrently prescribed medications — is itself an independent predictor of adverse drug events, hospitalization, and mortality, separate from any single drug’s inherent risk. Each additional medication adds interaction surface area, adherence complexity, and cumulative anticholinergic or sedative “burden.”
The medication list slider in this simulator models that reality directly: as list size grows, the probability of encountering a Beers-flagged agent grows with it, and the cumulative risk score compounds — not because any one drug is necessarily dangerous, but because the aggregate burden of many drugs interacting in an aging body is greater than the sum of its parts.
Each Beers Criteria table entry specifies a drug or drug class, the rationale for concern, a recommendation, and a quality-of-evidence / strength-of-recommendation rating. Automated and pharmacist-led medication reviews apply this table like a gate: every active medication on a patient’s list is checked against it, and any match is flagged for clinical reconsideration.
Every entry in the criteria follows a consistent structure so it can be applied consistently across settings:
• Organ System / Drug Class — e.g. CNS, anticholinergic, endocrine, cardiovascular • Drug(s) — the specific agent or class (e.g. diazepam, all benzodiazepines) • Rationale — the mechanism of harm (e.g. increased sensitivity, reduced clearance, higher risk of cognitive impairment, delirium, falls, fractures, motor vehicle crashes) • Recommendation — avoid, avoid in specific conditions, or use with caution • Quality of Evidence — high / moderate / low, graded by the expert panel • Strength of Recommendation — strong / weak
This structure is what lets a screening pass — whether performed by a clinical pharmacist, an EHR alert, or an automated tool like this simulator — flag a medication with a defensible, citable rationale rather than a vague “this seems risky.”
A handful of drug classes account for a disproportionate share of Beers-flagged prescriptions in practice:
• Benzodiazepines (diazepam, lorazepam, alprazolam) — all formulations flagged regardless of duration; older adults have increased sensitivity and decreased metabolism; associated with cognitive impairment, delirium, falls, fractures, and motor vehicle crashes • First-generation antihistamines (diphenhydramine, hydroxyzine, chlorpheniramine) — highly anticholinergic; often hidden in OTC sleep aids and cold remedies patients don’t report as “medications” • Anticholinergics broadly (tricyclic antidepressants, first-gen antipsychotics, some bladder antimuscarinics) — dry mouth, constipation, blurred vision, urinary retention, and — critically — cognitive impairment and delirium risk that compounds when multiple anticholinergics are stacked (the “anticholinergic burden”) • NSAIDs used chronically — GI bleeding/ulceration risk, acute kidney injury, and blood pressure elevation that can counteract antihypertensive therapy • Sulfonylureas, especially long-acting glyburide — prolonged, hard-to-reverse hypoglycemia in patients with reduced renal clearance • Long-term PPI use beyond 8 weeks without a clear indication — associated with C. difficile infection, bone fracture, and vitamin B12 deficiency with chronic use
A screening flag is a prompt for review, not an automatic stop order. A patient with metastatic cancer pain may appropriately continue an opioid despite a fall-risk flag; a patient with severe refractory anxiety may have already failed every non-benzodiazepine option. Good deprescribing practice, discussed in Stage 5, always pairs the automated flag with individualized clinical judgment, patient goals, and shared decision-making — which is precisely why the Beers Criteria describe themselves as decision support rather than a mandate.
The 2023 AGS Beers Criteria organize every flagged medication into one of five distinct tables, each with a different clinical logic. Understanding which bin a flagged drug falls into matters because it changes the recommended action — some drugs should simply be avoided, others are fine except in specific diseases, and others just need a lower dose or closer monitoring.
Drugs whose risk outweighs benefit for nearly all older adults regardless of diagnosis — benzodiazepines, first-generation antihistamines, tricyclic antidepressants, long-acting sulfonylureas, skeletal muscle relaxants, and barbiturates. These are the strongest recommendations in the document, generally backed by high-quality evidence and a strong recommendation to avoid.
Drugs that are reasonable in a healthy older adult but dangerous given specific comorbidities — NSAIDs in heart failure or chronic kidney disease, anticholinergics in dementia, alpha-blockers in a history of orthostatic hypotension, opioids or benzodiazepines in a history of falls or fractures. This table requires knowing the patient’s diagnosis list, not just their drug list.
A shorter list of drugs that aren’t contraindicated but warrant closer monitoring or dose limits — aspirin for primary prevention in patients over 70, long-term PPIs, and certain diuretics tied to hyponatremia risk. These are not "no" answers, they are "watch closely" answers.
Pairs or combinations of drugs that are each acceptable individually but dangerous together — opioids plus benzodiazepines (respiratory depression), warfarin plus NSAIDs (bleeding), and multiple CNS-active drugs stacked simultaneously (falls, delirium). This table was substantially expanded in the 2019 and 2023 updates as evidence linking combination therapy to adverse events matured.
Drugs cleared renally that accumulate to toxic levels as GFR declines — gabapentin, duloxetine, several antibiotics, and direct oral anticoagulants. Recommendations here are typically expressed as specific eGFR/CrCl thresholds requiring dose reduction or avoidance, making this the most quantitative of the five tables.
A single medication can sometimes appear in more than one table — for example, an NSAID appears in "avoid in conditions" for heart failure/CKD patients and separately drives drug-drug interaction risk when combined with an anticoagulant. Real-world screening tools must check every table, not stop at the first match.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Avoid in General | Benzodiazepines, 1st-gen antihistamines, TCAs | High risk regardless of diagnosis or context | Recommendation: avoid |
| Avoid in Conditions | NSAIDs in CKD/HF, anticholinergics in dementia | Risk depends on comorbid diagnosis | Recommendation: avoid if condition present |
| Use with Caution | Aspirin >70y, long-term PPIs | Net benefit uncertain; needs monitoring | Recommendation: monitor closely |
| Drug-Drug Interaction | Opioid + benzodiazepine, warfarin + NSAID | Combined effect exceeds sum of individual risk | Recommendation: avoid combination |
| Renal Dose Adjustment | Gabapentin, duloxetine, DOACs | Accumulates as eGFR/CrCl declines | Recommendation: dose-adjust by renal function |
No single PIM flag is, by itself, catastrophic — the danger is cumulative. Multiple studies link the total count and weighted burden of potentially inappropriate medications to a graded increase in falls, delirium, emergency visits, and hospitalization, in a dose-response relationship that mirrors what the rising burden tower in this simulator is meant to visualize.
Cohort studies consistently find that adverse drug event risk rises with the number of PIMs a patient is taking, not just their presence or absence. Patients on 2+ Beers-flagged medications show meaningfully higher rates of falls and hospitalization than those on a single flagged drug, and the relationship strengthens further at 3 or more. This is consistent with a cumulative-burden model: each additional CNS-active or anticholinergic drug adds independently to sedation, postural instability, and confusion risk, and the combined effect is often greater than any one drug’s individual contribution.
This is precisely the logic behind weighting each flagged drug class differently in a cumulative risk score rather than simply counting PIMs — a benzodiazepine (CNS-sedating, strong evidence of harm) should contribute more to the tower than a use-with-caution PPI, even though both are technically "flagged."
Falls are the most extensively documented consequence of PIM burden in older adults. Benzodiazepines, sedative-hypnotics, and anticholinergics impair postural control, reaction time, and alertness — all compounding fall risk that is already elevated by age-related sarcopenia and gait changes. A hip fracture following a fall carries roughly a 20-30% one-year mortality rate in adults over 80, making fall prevention one of the highest-yield targets of deprescribing.
Anticholinergic burden specifically correlates with delirium and long-term cognitive decline. Multiple anticholinergic drugs (a first-gen antihistamine plus a TCA plus an antimuscarinic bladder drug, for example) can each individually seem minor but sum to a "cumulative anticholinergic burden" score strongly associated with incident dementia in longitudinal studies — the same additive logic the burden tower visualizes.
A particularly insidious mechanism of harm is the prescribing cascade: a drug’s side effect is misdiagnosed as a new medical condition, and a second drug is prescribed to treat it — without ever questioning the first drug. Classic examples: an NSAID raises blood pressure, so an antihypertensive is added; a first-gen antihistamine causes urinary retention, so a bladder drug is added; an anticholinergic causes confusion, so an antipsychotic is added. Each new prescription adds its own PIM risk and interaction surface, and the cascade compounds the cumulative burden this stage visualizes — which is exactly why periodic, holistic medication review (not one-off reactive prescribing) is the core defense.
A landmark analysis found each additional inappropriate medication was independently associated with increased odds of an emergency department visit — underscoring that cumulative burden, not any single drug, is often the more clinically actionable target for intervention.
Identifying a PIM is only half the job. Deprescribing — the planned, supervised process of tapering or stopping a medication that is no longer beneficial or is now doing net harm — is the clinical action that converts a Beers Criteria flag into an actual improvement in patient safety. Done well, it is collaborative, gradual, and always paired with a safer alternative or a clear rationale for simply stopping.
Deprescribing is not simply "stopping drugs" — it is a structured clinical process:
• Prioritize — when multiple PIMs are present, target the highest-risk / lowest-benefit drug first (usually the highest-weighted "avoid in general" agent) • Taper, don’t abruptly stop — benzodiazepines, opioids, and some antidepressants require gradual dose reduction over weeks to months to avoid withdrawal or rebound symptoms • Substitute, don’t just subtract — replace a benzodiazepine for insomnia with cognitive behavioral therapy for insomnia (CBT-I); replace an NSAID for chronic pain with topical agents, acetaminophen, or physical therapy; replace a long-acting sulfonylurea with a shorter-acting or renally-safer agent • Monitor and reassess — track for withdrawal effects, symptom recurrence, and confirm the underlying condition doesn’t need the drug after all • Document and communicate — across all prescribers, since polypharmacy often originates from fragmented, uncoordinated care
The Beers Criteria are US-centric and only one of several explicit tools used internationally. STOPP/START (Screening Tool of Older Persons’ potentially inappropriate Prescriptions / Screening Tool to Alert doctors to Right Treatment), developed in Europe, differs by also flagging medication omissions — drugs that SHOULD be prescribed but aren’t (e.g. missing anticoagulation in atrial fibrillation). Comparative studies show STOPP/START and Beers agree on the majority of flags but each catches cases the other misses, and many geriatric pharmacists apply both.
Regardless of the tool, deprescribing succeeds only when paired with shared decision-making: explaining to the patient why a medication is being reconsidered, addressing their concerns (fear of symptom return, habituation), and respecting an informed patient’s choice to continue a flagged medication when the benefit genuinely outweighs the risk for them individually.
The AGS is explicit that the Beers Criteria are decision-support, not a mandate:
• They flag drug classes, not individuals — they cannot weigh a specific patient’s values, life expectancy, or goals of care • They can generate false positives — a hospice patient’s benzodiazepine for terminal anxiety is not truly "inappropriate" for their situation • They are a US-focused tool and may not reflect drug availability or practice patterns elsewhere • They don’t capture true "start" omissions the way STOPP/START does • Over-reliance on automated alerts contributes to "alert fatigue" among prescribers, potentially causing clinicians to override even valid flags
Used correctly — as a structured second opinion that prompts a conversation rather than a rule that ends one — the Beers Criteria remain one of the most effective tools available for reducing preventable medication harm in older adults.
AGS states plainly: "these criteria are not meant to be applied in a punitive manner." The goal of every flag in this simulator is the same as the goal of the real criteria — prompting a conversation about whether a medication still serves the patient, not issuing an automatic verdict.