Assessing decisional capacity when a psychiatric inpatient refuses recommended medication or hospitalization — autonomy first, protection when it is truly needed
Every legal and ethical framework governing medical decision-making starts from the same premise: adults are presumed capable of making their own treatment decisions. This is not a courtesy — it is a structural safeguard against paternalism. A psychiatric diagnosis, an inpatient admission, or an unconventional-sounding choice does not, by itself, rebut this presumption. Capacity must be affirmatively shown to be absent; it is never the patient's burden to prove they are competent.
Historically, psychiatric patients were often treated as though hospitalization itself suspended their right to make decisions. Modern mental health law and clinical ethics reject this: a person's status as a psychiatric inpatient carries no automatic implication about their decisional capacity for any specific choice.
The presumption of capacity does three things simultaneously:
1. It protects patients from having their autonomy stripped on the basis of diagnosis, appearance, or disagreement with staff. 2. It places the evidentiary burden on the clinical team, not the patient — incapacity must be demonstrated, not assumed. 3. It anchors the entire subsequent evaluation process: assessment begins as a genuine inquiry, not a search for confirmation of a foregone conclusion.
A patient who is floridly psychotic, severely depressed, or manic may still retain full or partial capacity for a given decision. Symptoms and capacity are correlated in the population but are not identical at the level of the individual patient — each must be evaluated on its own facts.
Capacity is not a single global switch. A patient may have capacity to refuse a blood draw but lack capacity, at the same moment, to weigh a complex decision about starting a new antipsychotic with significant metabolic risk. Capacity can also fluctuate: a patient in acute crisis may regain capacity within hours as agitation settles, intoxication clears, or delirium resolves.
Because of this specificity, "does this patient have capacity?" is not a well-formed clinical question. The well-formed question is: "does this patient have capacity to make this decision, about this treatment, right now?" Every stage of this simulator is anchored to that framing.
A treatment refusal becomes clinically significant not because it is a refusal, but because of what it risks and what it might indicate. Two circumstances typically prompt a formal evaluation: the refused treatment carries meaningful risk of serious harm if forgone, or there is independent clinical concern — confusion, disorganized thinking, inability to track a conversation — that reasoning may be impaired. Neither condition is met simply because a patient disagrees with the treatment team's recommendation.
The intensity of a capacity evaluation should be proportional to the consequences of getting it wrong — a principle sometimes called the "sliding scale" of capacity assessment. Refusing a low-risk, easily reversible intervention warrants a light-touch conversation. Refusing a treatment for a life-threatening condition, or refusing hospitalization while at imminent risk of serious harm to self or others, warrants a rigorous, well-documented, often multi-clinician evaluation.
This sliding scale is about the thoroughness of the assessment, not the standard for capacity itself. The definition of capacity does not change with risk — a patient facing a high-stakes decision is not held to a stricter bar simply because more is at stake. What changes is how carefully, and how many independent sources of information, the team uses to reach its determination.
Even a low-risk refusal can trigger assessment if the clinician independently observes signs suggesting impaired decision-making: disorganized or tangential speech that prevents coherent explanation of the choice, delirium, severe cognitive impairment, or a stated rationale that is internally contradictory or disconnected from the patient's actual circumstances (for example, denying the existence of a documented, observable illness in a way that is not merely a difference of opinion but a break from shared reality).
Critically, this concern must be observed independently of the refusal itself — the reasoning it triggers on is how the patient is thinking, not what they decided. A patient who refuses medication after a clear, coherent, well-reasoned explanation should not trigger this pathway merely because staff find the conclusion unwise.
The refusal is a prompt to look carefully — never a substitute for looking. Treating "refused recommended treatment" as itself diagnostic of incapacity collapses the very distinction this entire framework exists to protect.
The dominant clinical standard for capacity evaluation, widely used across psychiatry and consultation-liaison practice, breaks the assessment into four components (following Grisso and Appelbaum's influential framework). Each component is assessed specifically in relation to the treatment currently being refused — not as an abstract measure of intelligence, insight, or diagnosis.
Understanding asks whether the patient can absorb and accurately restate, in their own words, the basic factual information relevant to the decision: what the proposed treatment is, what condition it addresses, what the likely risks and benefits are, and what the reasonably available alternatives are (including the alternative of no treatment).
This is tested conversationally, not as a quiz — clinicians typically ask the patient to explain back, in their own words, what has been discussed. Poor understanding may stem from cognitive impairment, acute psychiatric symptoms disrupting attention, sensory or language barriers, or simply information delivered poorly — the last of these is not a capacity problem at all, and should prompt clearer communication before any determination is made.
Appreciation goes a level deeper than understanding: it asks whether the patient accepts that the facts genuinely apply to their own situation. A patient can accurately recite "antipsychotics can reduce hallucinations" (understanding) while insisting "but I don't have a psychiatric illness, so this doesn't apply to me, the voices are real recordings being played by neighbors" (a failure of appreciation).
Impaired appreciation is often driven by anosognosia (a neurologically or psychiatrically rooted lack of awareness of one's own illness) or by delusional beliefs that distort the patient's model of their own circumstances. This is distinct from a patient who understands they have a diagnosis and simply weighs the evidence differently, or who reasonably questions a diagnosis in a way any patient might.
Reasoning examines the process the patient uses to move from facts to a decision: can they compare treatment against non-treatment, weigh short-term discomfort against longer-term benefit, and follow a train of logical inference from premises to conclusion? Clinicians probe this by asking the patient to walk through their thinking — "what makes you choose this option over that one?"
The standard here is process, not conclusion. A patient who reasons "I understand the medication reduces relapse risk, but the side effects have been intolerable for me before and I would rather manage the risk of relapse than experience those side effects again" demonstrates intact reasoning even if the team disagrees with the ultimate weighting. A patient whose stated logic does not actually connect to their conclusion, or is driven by a delusional premise, demonstrates impaired reasoning.
The final component is the most basic: can the patient express a choice at all, and is that choice reasonably stable rather than shifting erratically from moment to moment in a way that reflects the instability of the decision-making process itself (rather than ordinary, healthy reconsideration)?
A patient who is mute, catatonic, or so severely disorganized that no consistent choice can be elicited fails this component regardless of how the other three might otherwise be assessed. Communication barriers unrelated to capacity — a stroke affecting speech, a language barrier, profound anxiety in the interview — should be addressed with accommodations (interpreters, alternative communication methods, a calmer setting) before concluding this component is impaired.
Weakest-link logic applies: a patient must clear the threshold on all four components to be found to have capacity for the decision at hand. A strong showing on three components does not offset a genuine failure on the fourth.
The single most common error in bedside capacity assessment is treating an unwise decision as evidence of incapacity. Capacity protects the right to make choices the clinical team would not make — including refusing medication, refusing hospitalization, or choosing an option with worse expected outcomes — provided the four-component process is intact. The moment "the patient disagrees with us" becomes indistinguishable from "the patient lacks capacity," the presumption of autonomy has effectively been erased.
Consider two patients who both refuse a recommended antipsychotic after a manic episode.
Patient A explains: "I understand it can help stabilize my mood and reduce the chance of another episode. I've been on this class of medication before and gained a significant amount of weight and felt sedated all day — for me, the risk of another manic episode, which I can usually catch early with my support system, is worth avoiding those side effects." This reasoning is coherent, connects facts to a conclusion, and reflects genuine values-based tradeoff. Patient A very likely retains capacity to refuse — even though the team may still recommend the medication and will document the disagreement.
Patient B explains: "I don't need it because I'm not sick, the hospital is testing me, and the pills are actually poison the doctors are being paid to give me." This is not a disagreement about tradeoffs; it reflects delusional beliefs that break appreciation and reasoning. Patient B is far more likely to lack capacity for this specific decision.
The refusal looks identical from the outside. The determination depends entirely on the process underneath it.
Clinicians are human, and there is a well-documented tendency to scrutinize refusals far more than acceptances — a patient who accepts a recommended treatment is rarely subjected to a formal capacity evaluation at all, even when their reasoning might be just as impaired. This asymmetry itself is worth naming, because it reveals that "capacity concern" can silently become a proxy for "disagrees with the team" if the process is not deliberately safeguarded.
Practical safeguards include: documenting the patient's reasoning in their own words rather than paraphrasing it as "irrational"; involving a second, ideally more neutral evaluator for high-stakes or contested cases; explicitly asking "would we be having this conversation if the patient had agreed with us?"; and revisiting the determination if initial impressions were formed before the structured four-component interview was actually completed.
Capacity assessment exists to distinguish "I disagree with your medical advice" from "I cannot meaningfully process this decision." Collapsing that distinction — even with good intentions — converts a protective evaluation into a tool for overriding autonomy whenever a patient says no.
When a structured evaluation concludes a patient genuinely lacks capacity for the decision at hand, the response is not simply to override the refusal. It follows a graded sequence that favors the option closest to what the patient would have chosen for themselves, resorting to more restrictive legal mechanisms only when necessary and only within the safeguards their jurisdiction provides.
If the patient completed a psychiatric advance directive, healthcare proxy designation, or similar document while they had capacity, and it speaks to the current situation, that document is generally the first and most authoritative source of guidance — it is the patient's own prior, capacitated voice. It should be honored unless circumstances have changed in ways the patient could not have anticipated, or the directive does not clearly address the situation at hand.
This step matters because it keeps the decision anchored in the patient's own values even when they can no longer express them in the moment — it is autonomy exercised in advance, not autonomy set aside.
Where no directive applies, a substitute decision-maker — a healthcare proxy, court-appointed guardian, or next-of-kin as defined by local law — is asked to decide using either the patient's own known wishes and values ("substituted judgment") or, where those are unknown, the patient's best interests. A good substitute decision-maker process actively solicits what is known about the patient's prior statements, values, and preferences rather than defaulting to whatever the clinical team originally recommended.
When no directive or substitute decision-maker is available or sufficient, and the situation is urgent — imminent risk of serious harm to the patient or others — clinicians may need to invoke emergency treatment provisions or formal involuntary treatment/commitment procedures. These vary substantially by jurisdiction but share common structural safeguards: independent medical certification, time limits, a right to legal representation, and judicial or independent tribunal review.
Even within this most restrictive pathway, the least-restrictive principle continues to apply at every decision point: the shortest necessary duration, the least invasive effective intervention, and ongoing reassessment for a return of capacity, at which point the patient's own voice resumes primacy.
Because capacity can fluctuate, a finding of incapacity is not a permanent status. Acute psychosis can respond to treatment, delirium can clear, intoxication can resolve, and mood episodes can stabilize — all of which can restore capacity for decisions the patient could not safely make hours or days earlier. Good practice builds in scheduled reassessment rather than treating an initial incapacity finding as settled for the remainder of an admission.
The entire pathway is ordered to intervene in the smallest way necessary to protect a patient who cannot currently protect themselves — never to make treatment more convenient for the care team. Every step up in restriction should be justified by the step below it being genuinely insufficient, not merely slower.