Shared decision-making simulator — balancing patient goals, medication benefit-harm evidence, and life expectancy
Stopping a medication a patient has taken for years is rarely a simple clinical calculation — it is a conversation. Shared decision-making (SDM) reframes deprescribing from a unilateral clinician decision into a collaborative process where evidence and patient values are deliberately brought together, spoken aloud, and reconciled.
Scott IA and colleagues (JAMA Internal Medicine, 2015) formalized deprescribing into a five-step clinical process:
1. Ascertain all medications the patient is currently taking, and the reasons for each 2. Consider overall risk of drug-induced harm to determine intensity of deprescribing intervention needed 3. Assess each drug for its current or future benefit versus current or future harm 4. Prioritize drugs for discontinuation — target those with least benefit / most harm, or greatest potential for withdrawal difficulty in reverse order 5. Implement and monitor a discontinuation regimen, with a plan for tapering, follow-up, and reinstatement if needed
The conversation itself is not a footnote to this process — it is embedded throughout. Step 3 in particular requires eliciting what "benefit" even means to this patient, which cannot be answered from a chart alone.
Elwyn et al. (2012, J Gen Intern Med) proposed the widely used "three-talk model" of SDM, adaptable directly to deprescribing conversations:
• Team talk — establish that a choice exists at all ("There may be a decision to make about this medication") and that the clinician wants to work through it together, not dictate it • Option talk — lay out the realistic alternatives (continue as-is, taper gradually, stop abruptly) with their trade-offs, avoiding jargon and using absolute risk framing rather than relative risk • Decision talk — explore what matters most to the patient, check understanding, and reach a decision that both parties can support, with an explicit review date
SDM is distinct from simple informed consent: it assumes genuine clinical equipoise (more than one reasonable option exists) and treats the patient as the expert on their own values while the clinician remains the expert on the evidence.
Deprescribing conversations routinely surface two anxieties that, left unaddressed, sink the discussion before evidence is even weighed:
• Fear of rebound or recurrence — "If I stop my statin, won't my cholesterol spike? If I stop my antidepressant, won't I relapse?" These fears are often valid for specific drug classes and warrant a concrete tapering and monitoring plan, not dismissal • Fear of abandonment — patients frequently interpret "let's stop this medication" as "I'm giving up on you" or "you don't think I'm worth treating anymore," especially in the context of reduced life expectancy
Reeve et al.'s Patients' Attitudes Towards Deprescribing (PATD) questionnaire found that most older adults are willing to stop a medicine if their doctor says it is possible — but only when the recommendation is framed as active, ongoing care rather than withdrawal of care.
The single most protective phrase in a deprescribing conversation is an explicit commitment to follow-up: "We are not stopping your care — we are changing the plan, and I will check in with you." Patients who hear a monitoring plan in the same breath as a stop recommendation report substantially higher trust and acceptance.
Clinical trials measure benefit in composite endpoints — mortality, major adverse events, hospitalization. Patients measure benefit in whether they can garden, remember their grandchildren's names, or avoid another fall. Eliciting and weighting goals of care is what allows a benefit-harm calculation to be personalized rather than generic.
Deprescribing frameworks typically organize patient priorities into four overlapping domains, each of which a given medication may serve differently:
• Symptom control — relief from pain, breathlessness, anxiety, or other distressing symptoms in the here-and-now • Longevity — extending overall survival or preventing a specific future event (stroke, fracture, cardiac event) • Independence / functional status — preserving the ability to perform daily activities, live at home, avoid falls or cognitive impairment caused by the drug itself • Quality of life — a broader, subjective sense of wellbeing that may trade off against any of the above (e.g., accepting shorter survival to avoid side effects)
A single medication, like a statin or a benzodiazepine, can score very differently across these four axes for the same patient — which is precisely why they must be elicited explicitly rather than assumed.
Fried, Bradley, Towle, and Allore (NEJM, 2002) surveyed seriously ill older adults and found that a majority would decline a treatment that extended life if it meant even moderate functional or cognitive burden — a striking reversal of the "more life is always better" assumption baked into many clinical guidelines.
This finding underlies a core deprescribing principle: guideline-recommended therapy calibrated to a 50-year-old's life expectancy and goals may be actively misaligned with an 85-year-old's stated priorities, even when the underlying disease risk calculation is unchanged. Goals of care are not static either — a patient recovering well from an acute illness may re-prioritize longevity, while one entering a period of functional decline may re-prioritize comfort and independence.
Practically, clinicians elicit goals with direct, open questions rather than assumptions:
• "What matters most to you in your health right now?" • "If you had to choose between living longer and feeling better day-to-day, which would you lean toward?" • "Are there things this medication or its side effects get in the way of?"
The answers are then used to weight the benefit-harm calculation for that specific medication: a patient who weights independence heavily will find a fall-risk-increasing sedative harder to justify even at low symptom burden, while a patient weighting longevity heavily may tolerate more side-effect burden from a disease-modifying therapy.
Goals of care elicitation is not a one-time checkbox. The same four-domain weighting should be revisited whenever health status changes materially — a new diagnosis, a fall, a hospitalization, or entry into a different phase of illness all justify re-asking the question.
Most preventive medications do not produce their benefit immediately — statins reduce cardiovascular events over years, bisphosphonates reduce fracture risk over years, cancer screening reduces mortality over a decade or more. The "time-to-benefit" concept asks a simple but often-skipped question: will this patient live long enough, in a state healthy enough, to actually collect the benefit?
Holmes, Hayley, Alexander, and Sachs (Archives of Internal Medicine, 2006) introduced "time-to-benefit" as a formal decision criterion: every preventive therapy has a minimum latency before its benefit exceeds the harm of taking it, derived from the time-to-separation of event curves in the pivotal trials.
When a patient's estimated life expectancy is shorter than a medication's time-to-benefit, that medication mathematically cannot deliver its intended benefit to this patient — only its near-term harms (side effects, cost, pill burden, drug interactions) remain in play. This reframes many "should we continue this?" conversations from a values question into an evidence question first, and a values question second.
Examples of published time-to-benefit estimates: statins for primary prevention (~2 years to first separation of event curves), bisphosphonates for fracture prevention (~1.5–3 years), and colorectal cancer screening (~10 years) all require the patient to survive well past the point of drug initiation before benefit accrues.
Several long-term medication classes have accumulated substantial evidence supporting safe deprescribing in appropriate patients:
• Proton pump inhibitors (PPIs) — frequently continued indefinitely after an initial short-term indication resolves; step-down or discontinuation is successful in the majority of patients without a strong ongoing indication (e.g., Barrett esophagus, severe erosive esophagitis) • Statins in limited life expectancy — for patients with a prognosis shorter than the drug's time-to-benefit, discontinuation trials (e.g., in advanced illness populations) have shown no significant mortality difference while improving quality of life and reducing pill burden • Benzodiazepines and sedative-hypnotics — long-term use carries accumulating fall, fracture, and cognitive-impairment risk that often outweighs the original symptom-control indication; gradual tapering protocols achieve discontinuation in a substantial proportion of long-term users
In each case, the harm side of the ratio (falls, cognitive effects, drug interactions, cost) is continuous and present from day one, while the benefit side may be deferred, uncertain, or already realized.
A working benefit-harm ratio for deprescribing combines several inputs, weighted by the patient's own goals from Stage 2:
• Magnitude of benefit — absolute risk reduction for the target outcome, not relative risk • Time-to-benefit versus estimated life expectancy — does the patient have enough runway to collect the benefit at all • Magnitude and immediacy of harm — side-effect burden, interaction risk, fall risk, cost, and administration burden, all of which accrue immediately regardless of remaining lifespan • Reversibility — how easily could the medication be restarted if stopping proves to be the wrong call
As time-to-benefit shrinks relative to estimated life expectancy, the ratio tips toward harm dominating; as patient preference for deprescribing rises, the practical threshold for tipping the decision drops further.
Time-to-benefit is a population-derived average from trial data, not a guarantee for an individual patient — it should inform, not replace, the individualized conversation about goals and preferences.
Once benefit, harm, time-to-benefit, and patient preference have been weighed together, the conversation converges on one of three pathways. The choice between them — and critically, the choice between tapering and abrupt discontinuation — carries its own evidence base and its own risks.
A common misconception frames deprescribing as a binary choice — keep the medication or stop it. The structured pathway instead recognizes three branches:
• Continue — the benefit-harm balance, time-to-benefit, and patient preference all favor keeping the medication as-is, possibly with dose or monitoring adjustments • Taper — the balance favors discontinuation, but the drug class carries meaningful withdrawal or rebound risk, so dose reduction proceeds gradually over weeks to months with scheduled check-ins • Stop — the balance favors discontinuation and the drug carries low withdrawal risk, allowing straightforward cessation (many PPIs, unnecessary supplements, drugs with resolved indications)
The branch chosen is not purely mechanical: a patient with a strong preference to stop but a drug with high withdrawal risk should still generally be routed to a taper branch, with the conversation reframed around timeline rather than whether to stop at all.
Several medication classes carry documented physiologic dependence and withdrawal syndromes that make abrupt discontinuation inappropriate regardless of how strongly the benefit-harm balance favors stopping:
• Benzodiazepines — abrupt cessation risks rebound anxiety, insomnia, and in severe/long-term use, seizures; structured tapers (commonly reducing by 10-25% every 1-2 weeks, slower near the end) achieve successful discontinuation in a majority of motivated long-term users • Opioids — abrupt discontinuation risks withdrawal symptoms and, paradoxically, increased illicit-use risk; gradual tapers with patient-paced flexibility are preferred • SSRIs/SNRIs — discontinuation syndrome (dizziness, "brain zaps," irritability) is common with abrupt stopping, particularly for shorter half-life agents; gradual dose reduction minimizes symptoms • Beta-blockers, clonidine — abrupt stopping risks rebound hypertension or tachycardia; tapering is standard of care
By contrast, many long-term PPIs, vitamins/supplements without indication, and drugs whose original indication has simply resolved can typically be stopped without a taper.
A structured decision pathway is not a one-way door. Every deprescribing plan should include an explicit statement that the medication can be restarted if symptoms recur, monitoring reveals a problem, or the patient changes their mind — this single reassurance measurably increases patient willingness to attempt a taper or stop in the first place.
Documenting the decision pathway (which branch, why, and the review date) also protects continuity of care: the next clinician who sees this patient should be able to see not just what was stopped, but the reasoning and the plan for reassessment.
Framing "stop" as "pause with a review date" rather than "stop forever" is one of the most consistently cited techniques for improving patient acceptance of deprescribing recommendations across the shared decision-making literature.
A deprescribing decision does not end the clinical relationship — it opens a monitoring period. Structured follow-up distinguishes safe, evidence-based deprescribing from simply discontinuing a drug and hoping for the best, and it is what allows patients and clinicians to trust the process enough to attempt it.
The Canadian Deprescribing Network (CDN), led by researchers including Barbara Farrell and Cara Tannenbaum, has published evidence-based deprescribing algorithms for several widely-overused drug classes — proton pump inhibitors (2017), benzodiazepine receptor agonists (2018), antipsychotics for behavioral symptoms of dementia, sulfonylureas and other diabetes medications, and cholinesterase inhibitors.
Each CDN algorithm follows a consistent structure: confirm the original indication and whether it still applies, assess candidacy for deprescribing, provide a stepwise tapering schedule where relevant, and specify what symptoms should trigger reassessment or reinstatement. This standardization gives clinicians a defensible, evidence-based script rather than an ad hoc decision made under time pressure.
A monitoring plan after a deprescribing decision typically tracks two distinct risks, which can be confused if not made explicit to the patient:
• Withdrawal or rebound effects — symptoms caused by physiologic adaptation to the drug itself (e.g., rebound acid hypersecretion after stopping a long-term PPI, rebound insomnia after stopping a benzodiazepine). These are usually time-limited and expected to resolve within days to weeks • Disease or symptom recurrence — return of the underlying condition the medication was treating (e.g., recurrent reflux, recurrent anxiety, rising blood pressure). This may indicate the medication should be reinstated rather than simply waited out
Distinguishing these two categories during a follow-up call or visit determines whether the correct response is reassurance and patience, or reinstatement of therapy.
A typical monitoring timeline after a taper or stop decision includes:
1. Early check-in (days to a few weeks) — by phone or portal message, focused on acute withdrawal or rebound symptoms 2. Interim check-in (during an active taper) — scheduled at each dose-reduction step to confirm tolerability before proceeding 3. Formal follow-up visit (weeks to a few months) — reassess the original indication, confirm no symptom recurrence, and formally close the loop on the decision 4. Longer-term surveillance — for drugs with delayed-recurrence risk, periodic reassessment continues at routine visits, with an explicit note in the record of what was stopped and why
This is the same infrastructure used for starting a new medication, applied symmetrically to stopping one — reinforcing to the patient that deprescribing receives the same clinical rigor and attention as prescribing.
A well-documented monitoring plan is what allows both clinician and patient to view a deprescribing decision as a supervised clinical trial of one, not an irreversible gamble — which is often the deciding factor in whether a patient agrees to attempt it at all.