Recognizing and treating confusion in the dying patient — CAM/Nu-DESC screening, reversible-cause triage, and goal-concordant symptom control near the end of life
Delirium is the single most common neuropsychiatric complication in the final weeks of life, yet it is missed in up to two-thirds of cases because clinicians expect agitation and instead find quiet withdrawal. Structured screening tools convert a vague clinical impression — "she seems different today" — into a reproducible, actionable diagnosis that changes management.
The CAM (Inouye et al., 1990) remains the most widely validated delirium screening instrument in medicine, adapted for palliative and hospice settings with minimal modification. It requires the presence of features 1 and 2, plus either 3 or 4:
1. Acute onset and fluctuating course — a change from the patient's baseline mental status that waxes and wanes over hours, obtained from a caregiver or the chart, not just the patient 2. Inattention — difficulty focusing, easily distracted, or struggling to keep track of conversation; tested informally by asking the patient to recite the months of the year backward or spell a word backward 3. Disorganized thinking — rambling or irrelevant conversation, unclear or illogical flow of ideas, unpredictable switching from subject to subject 4. Altered level of consciousness — anything other than "alert": vigilant (hyperalert), lethargic (drowsy, easily aroused), stuporous (difficult to arouse), or comatose
A positive CAM (features 1 + 2 + [3 or 4]) has sensitivity of 94% and specificity of 89% for delirium as diagnosed by a geriatric psychiatrist reference standard — strong enough to be used as a rule-in and rule-out test in most clinical settings. The exam takes under five minutes and can be taught to bedside nurses, hospice aides, and family caregivers with brief training.
CAM asks about acute change from baseline — which means the single most valuable screening question a clinician can ask a family member is simply: "Is this how they usually are, or is this different?" A caregiver who says "she was joking with me yesterday and today she doesn't know where she is" has just delivered a positive CAM feature 1 without any formal tool at all.
Many patients approaching death are too somnolent, dyspneic, or weak to participate in an interactive cognitive test. The Nursing Delirium Screening Scale (Nu-DESC; Gaudreau et al., 2005) solves this by relying entirely on brief observation, making it especially well suited to hospice and inpatient palliative units where nursing staff have continuous contact.
Nu-DESC rates five items, each scored 0 (absent) to 2 (severe), observed over a nursing shift:
• Disorientation — verbal or behavioral manifestation of not being oriented to place, time, or misidentifying persons • Inappropriate behavior — behavior inappropriate to place and/or for the person (pulling at tubes, trying to get out of bed when this is unsafe or not indicated) • Inappropriate communication — communication inappropriate to place and/or for the person (incoherent, unrelated, or unintelligible speech) • Illusions/hallucinations — seeing or hearing things that are not there, distorting actual stimuli • Psychomotor retardation — delayed responsiveness, few or no spontaneous actions/words, e.g. when patient is prodded, reaction is deferred and/or patient is unarousable
A total score ≥2 across the five items flags probable delirium, with sensitivity around 86% and specificity around 87% against DSM criteria. Because it requires no direct patient interrogation, Nu-DESC can be completed even in obtunded, actively dying patients and repeated every shift to track trajectory.
Delirium is classified by psychomotor presentation, and the distribution is heavily skewed toward the subtype clinicians recognize least reliably:
Hypoactive delirium (50–70% of cases): lethargy, reduced responsiveness, withdrawal, decreased speech, appearing to sleep more. This is frequently mistaken for expected end-of-life fatigue, depression, or simply "the dying process," and is the subtype most often missed entirely — one study found hypoactive delirium was recognized by nursing staff in fewer than 20% of confirmed cases.
Hyperactive delirium (10–15% of cases): agitation, restlessness, hypervigilance, hallucinations, combativeness, pulling at lines. This is the subtype patients, families, and staff instinctively recognize as "delirium" or "confusion" because it is behaviorally dramatic — but it represents a minority of episodes.
Mixed delirium (30–40% of cases): fluctuation between hypoactive and hyperactive states, sometimes within the same day — calm and withdrawn in the morning, agitated and hallucinating by evening (a pattern sometimes called "sundowning" when it clusters in late afternoon/evening).
Subtype matters clinically: hyperactive and mixed delirium are more likely to prompt pharmacologic intervention because they cause visible distress and safety risk, while hypoactive delirium — despite being just as strongly associated with underlying physiologic derangement and equally distressing to experience — is under-treated because it looks peaceful from the outside.
Once delirium is confirmed, the pivotal clinical question is not "what is causing this?" in the abstract, but "how much workup is proportionate given this patient's prognosis and goals?" A patient with weeks to live and a single clear trigger deserves active investigation; a patient actively dying over hours does not need a urine culture. Reversible-cause assessment is triage, not an exhaustive checklist.
Delirium is almost always multifactorial — a vulnerable brain (advanced illness, prior cognitive impairment, older age) tipped over threshold by one or more precipitating insults. Common reversible or partially reversible triggers in palliative populations, roughly in order of how often a focused work-up finds them:
• Opioid neurotoxicity — accumulation of active metabolites (e.g., morphine-3-glucuronide) especially with renal impairment, dose escalation, or dehydration; presents with agitation, myoclonus, hyperalgesia, and vivid hallucinations. Often mistaken for "worsening pain" prompting further opioid escalation, which paradoxically worsens the picture — a dangerous feedback loop • Urinary retention and fecal impaction — extremely common, easily missed, and easily fixed; a bladder scan and rectal exam belong in almost every delirium work-up regardless of prognosis • Dehydration and electrolyte derangement — hyponatremia, hypernatremia, and volume depletion from reduced oral intake, vomiting, or diuretics • Hypercalcemia of malignancy — seen in up to 30% of cancer patients at some point, causes lethargy, confusion, constipation, and polyuria; readily treated with IV fluids and bisphosphonates/denosumab if consistent with goals • Infection — urinary tract infection and pneumonia are the most frequent infectious triggers; fever may be blunted or absent in frail, elderly, or actively dying patients • Hypoxia and metabolic derangement — hepatic or renal failure, hypoglycemia • Medication effects beyond opioids — benzodiazepines, anticholinergics, corticosteroids, and polypharmacy interactions • Uncontrolled pain or unmet physical needs — pain itself is a potent delirium trigger; the agitated patient may be in pain, not merely confused
The intensity of reversible-cause assessment should be explicitly calibrated to prognosis and to the patient's previously expressed goals — a principle sometimes summarized as "treat what is treatable and consistent with the goals of care, and no more":
Prognosis of weeks to months: a fuller work-up is usually appropriate — basic metabolic panel, calcium, urinalysis, medication reconciliation with attention to opioids and anticholinergics, bladder scan, and consideration of infection work-up if clinically indicated and consistent with goals. Correcting a single clear trigger here can restore meaningful lucidity for remaining time that matters greatly to patients and families (a last coherent conversation, presence at a family event).
Prognosis of days: work-up narrows to bedside, low-burden interventions — bladder scan and straight catheterization if retention suspected, opioid rotation or dose reduction if neurotoxicity suspected, disimpaction if indicated. Blood draws and imaging are reconsidered against burden.
Prognosis of hours: formal investigation is generally not indicated. The clinical priority shifts entirely to comfort, and the delirium is treated symptomatically as part of the active dying process (see Stage 5).
This calibration must be explicit and discussed with the interdisciplinary team and family — otherwise default hospital reflexes (drawing labs, ordering imaging) continue by inertia even when they no longer serve the patient.
A practical bedside heuristic used in many palliative programs: ask "if we found and reversed one thing right now, would it change what this patient experiences in the time they have left?" If the honest answer is no — because death is imminent regardless — the work-up should stop and management should shift entirely to comfort-focused symptom control.
Multiple prospective studies of dying inpatients (Lawlor et al., Archives of Internal Medicine, 2000, among the most cited) found that while roughly half of delirium episodes earlier in a terminal illness trajectory are reversible with directed treatment, that fraction falls sharply — to well under 20% — in patients within the final 24 to 48 hours of life. The etiology in this window is typically multi-organ failure occurring simultaneously: renal and hepatic clearance failing together, progressive hypoxia, cytokine-mediated encephalopathy, and irreversible structural changes, none of which is correctable by any single intervention.
This is the biological basis for the clinical concept of terminal restlessness or terminal agitation discussed in Stage 5: it is not that clinicians have failed to find the cause, but that the cause is the dying process itself, occurring on every organ system at once. Recognizing this shift — from "delirium as a treatable complication" to "agitation as an expected feature of active dying" — is one of the most important judgment calls in end-of-life care, because it reorients the entire care plan from diagnostic pursuit to comfort and family support.
Non-pharmacological strategies are first-line for mild-to-moderate delirium and remain the essential foundation of care even after medications are started — drugs manage symptoms, but environment and human presence address the vulnerable, overwhelmed brain generating them. These interventions carry no risk of oversedation and are often more effective than clinicians expect.
A consistent, low-stimulation, orienting environment reduces both the incidence and severity of delirium and is recommended by every major palliative care guideline before or alongside pharmacologic treatment:
• Reorientation — brief, gentle, repeated reminders of who is present, where the patient is, and what is happening ("It's Tuesday afternoon, you're at home, I'm your daughter Anna, and the nurse is here to check on you"). Avoid confrontational correction of fixed false beliefs, which increases distress without benefit • Lighting and day–night cues — natural light during the day, dim lighting and quiet at night, visible clocks and calendars, to preserve circadian structure and reduce "sundowning" • Minimize room changes and staff turnover — consistent caregivers (same family members, same hospice aide) reduce the disorientation of unfamiliar faces • Reduce noise and unnecessary stimulation — alarms, overhead pages, television left on, unnecessary vital sign checks during sleep hours • Correct sensory deficits — ensure glasses, hearing aids, and dentures are in place; uncorrected sensory impairment is a frequently overlooked delirium amplifier • Familiar objects and photographs — a familiar blanket, photo, or piece of music can be grounding • Safety measures — fall precautions, low bed height, avoidance of physical restraints (which increase agitation and injury risk rather than reducing them)
Family presence is simultaneously a therapeutic intervention and a source of profound distress that itself needs active management. Practical family-facing interventions include:
• Explaining delirium explicitly as a medical syndrome — families who understand that agitation or incoherent speech represents a brain affected by illness, not the "real" person or a psychiatric breakdown, cope better and can participate more effectively in reorientation • Coaching in calm, simple communication — short sentences, one idea at a time, avoiding open-ended or testing questions ("Do you know where you are?") which increase distress when the answer is wrong • Involving family in reorientation and comfort measures — holding a hand, familiar voice and touch, playing preferred music, can be more calming than any medication • Preparing family for fluctuation — explaining that lucid intervals will occur and can be treasured, but do not mean the delirium has resolved, prevents false hope followed by repeated grief • Protecting family from exhaustion — rotating presence, permission to rest, chaplaincy and social work support, since sustained bedside vigilance during a loved one's delirium is itself a recognized risk factor for complicated grief
A careful medication review is one of the highest-yield, lowest-risk interventions available and is frequently the step most often skipped in busy clinical settings:
High-risk deliriogenic drug classes to review and, where possible, taper or discontinue: • Anticholinergics — diphenhydramine, hydroxyzine, oxybutynin, and many older antiemetics and sleep aids carry substantial anticholinergic burden and are strongly associated with delirium • Benzodiazepines — paradoxically can precipitate or worsen delirium in most patients (with important exceptions covered in Stage 4), particularly in opioid-naive or elderly patients • Corticosteroids — can cause steroid-induced psychosis or agitation, especially at higher doses • Opioids — not stopped, but doses and formulations reviewed for neurotoxic metabolite accumulation, with rotation considered (see Stage 2) • Any medication no longer serving an active symptom goal — statins, vitamins, and other long-term preventive medications with no near-term benefit are routinely discontinued in the terminal phase ("deprescribing"), simplifying the regimen, reducing pill burden, and removing potential contributors
Deprescribing decisions should be explained to families as an active, positive step — "we are removing medicines that are no longer helping and may be contributing to the confusion" — rather than allowed to be perceived as "giving up," which is a common and important source of family distress if left unaddressed.
When distressing symptoms persist despite non-pharmacological measures, or when agitation poses a safety risk, low-dose antipsychotics are the pharmacologic mainstay for hyperactive and mixed delirium. The evidence base here is more nuanced — and in some respects more cautionary — than older teaching suggests, and benzodiazepines are reserved for specific, well-defined indications rather than used as general sedatives.
Haloperidol remains the most extensively studied and widely used first-line agent for distressing delirium symptoms (hallucinations, paranoia, agitation) in palliative care, valued for its predictable pharmacology, multiple routes of administration, and relative lack of anticholinergic burden compared with other antipsychotics:
Haloperidol: • Typical starting dose: 0.5–1 mg orally, subcutaneously, or intravenously, repeated every 1–4 hours as needed, with cautious upward titration guided by response • Low doses target psychotic and agitated symptoms with less sedation than higher doses; higher doses (used for severe refractory agitation) increase risk of extrapyramidal side effects and QT prolongation • Subcutaneous administration is especially useful in the home hospice setting where IV access is unavailable and oral intake has stopped
Atypical antipsychotics (alternatives or adjuncts): • Olanzapine (2.5–5 mg PO/SL/ODT daily-BID) — more sedating than haloperidol, useful when sedation is also desired, available as an orally disintegrating tablet for patients with swallowing difficulty • Risperidone (0.25–0.5 mg PO BID) — lower sedation profile, oral solution available • Quetiapine (12.5–25 mg PO, more sedating, sometimes preferred at night for sleep-wake reversal) • All antipsychotics carry a boxed warning for increased mortality in elderly patients with dementia-related psychosis; this warning must be weighed against goals of care and symptom burden in the terminal setting, where comfort is typically the paramount consideration
Monitoring: baseline and periodic assessment for extrapyramidal symptoms, QTc prolongation risk (particularly with concurrent QT-prolonging drugs), and response using a validated scale (CAM severity, Richmond Agitation-Sedation Scale — RASS) to titrate rather than escalate blindly.
For decades, haloperidol was taught almost reflexively as the default treatment for any delirium symptom. A landmark 2017 randomized, double-blind, placebo-controlled trial (Agar et al., JAMA Internal Medicine) in palliative care inpatients complicated this picture substantially:
• The trial compared haloperidol, risperidone, and placebo for mild-to-moderate delirium symptoms in hospice/palliative populations • Both antipsychotics were associated with WORSE delirium symptom scores than placebo, more extrapyramidal effects, and — in the haloperidol arm — higher mortality • Non-pharmacological management alone outperformed antipsychotics for mild-to-moderate symptom severity in this population
This trial reshaped guideline recommendations (including NCCN and multiple palliative care society statements): antipsychotics are now generally reserved for more severe, distressing symptoms — significant hallucinations, paranoia causing fear, or agitation posing safety risk — rather than used routinely for mild confusion. For mild delirium without associated distress, non-pharmacologic measures alone, with close monitoring, are now preferred as first-line management.
This does not mean antipsychotics have no role — for severe, distressing hyperactive symptoms unresponsive to environmental measures, they remain the most evidence-supported pharmacologic option — but it does mean the decision to medicate should be driven by symptom severity and patient distress, not by the mere presence of confusion.
Key Insight: The 2017 Agar et al. RCT reversed decades of reflexive haloperidol prescribing for delirium of any severity. The modern standard is to reserve antipsychotics for moderate-to-severe, distressing symptoms — hallucinations causing fear, agitation risking safety — while managing mild delirium with non-pharmacological care and close observation alone, since drug treatment in mild cases was shown to do more harm than good.
Benzodiazepines are not first-line for typical delirium because, in most patients, they can worsen confusion, disinhibition, and paradoxical agitation, and are associated with delirium as a precipitating cause in their own right (see Stage 3). Their use in the delirium context is deliberately restricted to specific circumstances where their pharmacology is uniquely indicated:
1. Alcohol or benzodiazepine withdrawal — delirium tremens and sedative-hypnotic withdrawal require benzodiazepines (e.g., lorazepam) as definitive, disease-specific treatment; antipsychotics alone can lower seizure threshold and are inadequate 2. Seizure-related agitation or known seizure disorder — benzodiazepines have anticonvulsant activity that antipsychotics lack 3. Agitation refractory to antipsychotics — when adequately dosed haloperidol or an atypical antipsychotic fails to control severe, dangerous agitation, a benzodiazepine (commonly lorazepam or midazolam) may be added as an adjunct, often in combination with rather than instead of the antipsychotic 4. Terminal restlessness in the actively dying patient where distinguishing treatable delirium from the dying process is no longer the priority and comfort is paramount — here benzodiazepines (particularly subcutaneous or intravenous midazolam) are commonly used, sometimes as part of proportionate palliative sedation (see Stage 5)
Outside these situations, benzodiazepines used as a general-purpose sedative for delirium risk worsening the underlying confusional state, increasing fall risk, and masking the ability to reassess mental status — which is why careful indication-matching, not blanket avoidance or blanket use, is the guiding principle.
In the final day or two of life, agitation frequently represents terminal restlessness — a manifestation of the dying process itself rather than a reversible medical complication — and the clinical task shifts from diagnosis to comfort, presence, and honest conversation with the family about what is happening and why further investigation is no longer the priority.
Terminal restlessness (also called terminal agitation) describes agitated, restless behavior — picking at bedsheets, moaning, grimacing, trying to climb out of bed, calling out — that emerges in the final 24 to 48 hours of life, when the etiology is the simultaneous, irreversible failure of multiple organ systems rather than any single correctable trigger.
Clinical features that favor terminal restlessness over reversible delirium include: • Clear trajectory of decline already established, with other signs of active dying present (mottling, altered breathing patterns such as Cheyne-Stokes respiration, decreased urine output, loss of ability to swallow) • Absence of a single obvious reversible trigger despite reasonable bedside assessment (bladder not distended, no clear new infection, medications already reviewed) • Prior discussion and clinical consensus that the patient is in the actively dying phase, typically hours to 1–2 days
This distinction matters enormously for the care plan: a patient with weeks of prognosis and new agitation deserves the reversible-cause work-up detailed in Stage 2, while a patient who is actively dying with the same behavioral presentation is best served by prompt comfort-focused symptomatic treatment, because further investigation would add burden without meaningfully changing the outcome or the patient's remaining experience.
Importantly, this is a clinical judgment made by the interdisciplinary team, not an excuse to skip assessment — bladder scanning and a basic review for obvious, low-burden fixable causes remain appropriate even in the actively dying patient, since a full bladder is trivially correctable and causes real suffering.
How clinicians communicate about terminal restlessness shapes family memory of the death for years afterward, making this one of the highest-stakes conversations in palliative care. Effective communication generally includes:
• Naming what is happening plainly — "Your mother's body is now in the process of dying, and this kind of restlessness is a common part of that process. It does not mean she is in pain the way we usually think about pain, and it does not mean we have missed something we could fix." • Explaining the shift in focus — from looking for a cause to treating the symptom directly, and framing this explicitly as continued active care rather than withdrawal of care • Preparing families for what they may see — periods of moaning or grimacing that may look distressing but, once medications are adjusted, are managed; changes in breathing pattern; skin color changes; decreasing responsiveness • Correcting common misconceptions — that morphine or sedatives given now are "hastening death" (addressed via the doctrine of double effect: medication given to relieve suffering, at doses proportionate to the symptom, is ethically and legally distinct from euthanasia even if it may have a secondary effect on the timing of death), or that a calm-appearing patient does not need medication because they seem peaceful • Offering concrete roles — holding a hand, talking or reading aloud, playing music — so family members have something meaningful to do rather than only watch • Revisiting the conversation as things change — a single conversation is rarely sufficient; short, repeated check-ins as the trajectory evolves reduce cumulative distress
Families overwhelmingly report, in bereavement research, that clear and repeated explanation of what agitation means and why the care plan is what it is — even when nothing can "fix" the underlying process — reduces long-term distress and complicated grief more than any single medication choice. Communication is a core clinical intervention in this stage, not a soft add-on to the "real" treatment.
For the roughly 10–15% of dying patients whose distress — physical or existential — remains severe and refractory despite optimized non-pharmacological care and standard pharmacologic management, palliative sedation becomes an appropriate option, governed by well-established ethical and clinical principles:
• Proportionality — sedation is titrated to the minimum level that relieves the specific refractory symptom, not automatically to unconsciousness; many patients achieve comfort with moderate sedation while remaining arousable • Refractoriness must be established — the symptom must have failed to respond to appropriate trials of standard therapy (correct antipsychotic dosing, opioid titration for pain, other targeted measures) delivered within a timeframe and by a route tolerable to the patient, with a clinical consensus (often including an interdisciplinary or ethics consultation for complex cases) that no other reasonable intervention exists • Common agents — midazolam (subcutaneous or intravenous infusion, titrated) is the most frequently used agent for palliative sedation because of its rapid onset, short half-life, and ease of titration; other options include lorazepam or, in specialized settings, phenobarbital or propofol • Distinct from euthanasia — the intent of palliative sedation is relief of refractory suffering, with dose titrated to symptom control; it is ethically and legally distinguished from euthanasia (whose intent is to cause death) by intent, dose proportionality, and the doctrine of double effect, and is endorsed by major palliative care and ethics organizations as a legitimate end-of-life intervention • Family and team communication — the decision, its rationale, expected effects, and what family should anticipate are discussed explicitly beforehand, since sedation changes how the family experiences the final hours and this transition deserves the same careful preparation described above
Palliative sedation is used in a minority of dying patients — most terminal restlessness is well controlled with careful non-pharmacological care plus low-dose antipsychotic and/or benzodiazepine titration — but knowing when and how to escalate to it is an essential competency for anyone caring for dying patients.