ICU delirium prevention and management — CAM-ICU, ICDSC, ABCDEF bundle, dexmedetomidine evidence

1. Overview and Clinical Importance

Delirium is the most common form of acute brain dysfunction encountered in the ICU, affecting 20-80% of mechanically ventilated patients depending on the population studied and the screening tool applied. It is not a nuisance; it is an independent predictor of prolonged mechanical ventilation, longer ICU and hospital stay, increased mortality, and long-term cognitive impairment. A single day of delirium adds approximately one day to ICU length of stay. The cognitive sequelae, sometimes grouped under the banner of post-intensive care syndrome (PICS), can persist for years.

Despite its prevalence and prognostic weight, delirium remains under-recognised when clinicians rely on clinical gestalt rather than validated screening tools. The ANZ ICU community has increasingly embedded systematic screening into daily workflow, and the ABCDEF bundle represents the dominant organisational framework for both prevention and management.

This note covers:

• Neurobiological mechanisms underpinning delirium
• Validated screening instruments (CAM-ICU, ICDSC)
• The ABCDEF bundle in full
• Non-pharmacological prevention strategies
• Pharmacological management, dexmedetomidine, haloperidol, quetiapine, and the evidence base from SPICE-III, MIND-USA, and AID-ICU
• Avoidance of benzodiazepines
• Practical ANZ application including dosing, monitoring, and decision triggers

2. Physiological and Pharmacological Framework

2.1 Neurobiological Basis of Delirium

ICU delirium is best understood as a failure of cerebral homeostasis driven by multiple converging insults. No single mechanism is sufficient; the final common pathway is disrupted neuronal signalling manifest as global cognitive dysfunction with fluctuating arousal, attention deficits, and disorganised thinking.

Key mechanisms:

Neuroinflammation. Systemic inflammation, from sepsis, trauma, surgery, or any critical illness, activates brain-resident microglia via several routes: direct cytokine crossing of a disrupted blood-brain barrier (BBB), vagal afferent signalling, and circumventricular organs that lack a complete BBB. Activated microglia release pro-inflammatory mediators (TNF-α, IL-1β, IL-6) that impair synaptic transmission and injure neurons, particularly cholinergic neurons of the basal forebrain, which are critical for attention and arousal.

Cholinergic deficiency. The cholinergic hypothesis proposes that relative ACh deficiency (from neuroinflammation-mediated neuron loss, substrate depletion, or exogenous anticholinergic drug burden) disrupts the ascending reticular activating system (ARAS) and cortical attention networks. This explains the hyperactive, hypoactive, and mixed subtypes as different points on a spectrum of ARAS dysfunction.

Dopaminergic excess. Dopamine excess in the mesolimbic and mesocortical tracts is implicated in the perceptual disturbances and psychomotor agitation of hyperactive delirium. This provides the pharmacodynamic rationale for dopamine D2 receptor antagonism, the mechanism of both haloperidol and quetiapine.

GABAergic dysregulation. Benzodiazepines positively modulate GABA-A receptors, producing sedation; however, paradoxically, GABA excess (or sudden withdrawal) is strongly linked to delirium. The widespread thalamo-cortical network suppression produced by benzodiazepines likely disrupts the brain's default mode network and frontal executive circuits, precipitating delirium. Benzodiazepine-associated delirium is dose-dependent and consistent across multiple observational and randomised datasets.

Serotonergic and noradrenergic contributions. Altered serotonin signalling (particularly serotonin excess) may underlie agitated delirium subtypes. The locus coeruleus, the primary source of noradrenergic tone, modulating arousal and attention, is highly sensitive to hypoxia, hypoperfusion, and inflammatory cytokines. Alpha-2 agonists such as dexmedetomidine act here: by stimulating presynaptic α-2 receptors in the locus coeruleus, they reduce noradrenergic firing, producing a form of sedation more closely resembling natural sleep than GABA-mediated sedation, with preservation of rousability and without marked respiratory depression.

Sleep-wake disruption. Normal sleep architecture, particularly slow-wave (N3) sleep, is required for synaptic homeostasis, amyloid clearance via the glymphatic system, and emotional memory consolidation. ICU-acquired sleep fragmentation (from noise, nursing interventions, continuous lighting, and sedative-analgesic regimens that suppress N3 sleep) both predisposes to and perpetuates delirium. Melatonin secretion is severely blunted in critically ill patients, disrupting circadian entrainment.

Metabolic and haemodynamic vulnerabilities. Hypoxia, hypercapnia, hepatic encephalopathy, uraemia, electrolyte disturbance (particularly hyponatraemia, hypercalcaemia), and hypoglycaemia all impair neuronal energy supply or neurotransmitter synthesis, with delirium as a common final pathway.

2.2 Pharmacological Framework for Treatment Agents

3. Screening for ICU Delirium

Systematic, twice-daily screening is the standard of care. Two validated tools are endorsed by the PADIS (Pain, Agitation/Sedation, Delirium, Immobility, Sleep) guidelines and are widely used in ANZ ICUs.

3.1 CAM-ICU (Confusion Assessment Method for the ICU)

The CAM-ICU was adapted from the original CAM (1990) by Ely and colleagues for use in non-verbal, mechanically ventilated patients. It is a binary (delirium present/absent) assessment.

Four features assessed:

1. Acute onset or fluctuating course, change from baseline mental status, or fluctuation in the last 24 hours (yes/no based on nursing observation or family report)
2. Inattention, assessed using the "Attention Screening Examination" (ASE): the patient squeezes a hand when they hear the letter "A" in a 10-letter sequence (SAVEAHAART), or a picture recognition task. Score ≥ 2 errors = inattention present
3. Altered level of consciousness, current RASS other than 0 (alert and calm); or if RASS −3 to −5, the patient is unarousable and cannot be assessed (RASS −4/−5 = unable to assess)
4. Disorganised thinking, four yes/no questions + a two-step command (e.g. hold up this many fingers, then switch hands). ≥ 2 errors = disorganised thinking present

Delirium = Features 1 + 2 + either 3 or 4 present.

Sensitivity ~80%, specificity ~96% against psychiatrist DSM diagnosis. Requires RASS ≥ −3 for assessment; at RASS −4/−5 the brain is too suppressed to assess delirium (though deep coma may represent an extreme form of the same pathological spectrum).

Key limitation: hypoactive delirium is easily missed without systematic screening; the hyperactive phenotype (agitation, pulling lines) is obvious, but hypoactive delirium (quiet withdrawal, flat affect, staring) is equally prognostically serious and frequently undetected.

3.2 ICDSC (Intensive Care Delirium Screening Checklist)

Developed by Bergeron and colleagues, the ICDSC is an 8-item checklist completed by nursing staff over an 8-12 hour shift, scoring observed behaviours rather than active bedside assessment.

Eight items (each scores 0 or 1):

1. Altered level of consciousness
2. Inattention
3. Disorientation
4. Hallucinations, delusions, or psychosis
5. Psychomotor agitation or retardation
6. Inappropriate speech or mood
7. Sleep-wake cycle disturbance
8. Symptom fluctuation

Scoring: ≥ 4 = delirium (sensitivity ~99%, specificity ~64%); scores 1-3 = subsyndromal delirium.

The ICDSC has higher sensitivity but lower specificity than CAM-ICU. It captures subsyndromal delirium, a state with intermediate outcomes between no delirium and full delirium, that CAM-ICU may miss. The ICDSC may be more practical in busy nursing environments as it integrates naturally into shift documentation.

In ANZ practice, both tools are used in different units. CAM-ICU is more commonly implemented as the primary tool, often integrated into electronic medical records with RASS assessment as a prerequisite. Either tool is acceptable for fellowship examination purposes; you should know both in detail.

4. The ABCDEF Bundle

The ABCDEF bundle (sometimes called the ICU Liberation Bundle) is the dominant evidence-based framework for ICU delirium prevention and, indeed, for optimising broader ICU outcomes. It integrates previously siloed interventions into a coordinated daily workflow. The large observational ABCDEF Bundle Collaborative (Pun et al. 2019, >15,000 patient-days) demonstrated dose-dependent improvements across all outcomes with higher bundle compliance.

A, Assess, Prevent, and Manage Pain

Pain is an independent precipitant of delirium. Undertreated pain activates the HPA axis and sympathetic nervous system, worsening neuroinflammation and sleep disruption. Over-treatment with opioids (excessive sedation) suppresses arousal and can paradoxically precipitate delirium.

• Assess pain regularly using validated tools: NRS (numeric rating scale) for verbal patients; CPOT (Critical Care Pain Observation Tool) or BPS (Behavioural Pain Scale) for non-verbal/mechanically ventilated patients
• Analgesia-first ("analgosedation") approach: treat pain before adding sedation
• Multimodal analgesia, paracetamol, ketamine, regional techniques, NSAIDs (selected patients), reduces opioid requirements and therefore reduces opioid-related delirium

B, Both Spontaneous Awakening Trials and Spontaneous Breathing Trials

The "wake up and breathe" protocol, validated in the AWAKENING AND BREATHING CONTROLLED (ABC) trial, pairs daily Spontaneous Awakening Trials (SAT) with Spontaneous Breathing Trials (SBT).

• SAT: Nurse-driven cessation of sedation and/or analgesic infusion for a set period. Enables delirium assessment, neurological evaluation, and degree of sedation titration
• SBT: Respiratory therapist/nurse-driven trial of unassisted or minimally supported breathing to assess readiness for extubation
• Combined SAT + SBT reduced ventilator days, ICU stay, and 1-year mortality in the ABC trial compared to SBT-only

C, Choice of Analgesia and Sedation

• Target the lightest sedation depth consistent with patient safety and comfort: RASS −1 to 0 as a default target
• Propofol or dexmedetomidine preferred over benzodiazepines for sedation (see Section 6)
• Consider protocolised sedation (nurse-driven titration), validated in SEDCOM trial (dexmedetomidine vs midazolam) and MENDS2 (propofol vs dexmedetomidine in sepsis)
• Regularly reassess sedation need: can the patient be managed without a continuous sedation infusion?

D, Delirium Assessment and Management

• Twice-daily systematic delirium screening (CAM-ICU or ICDSC)
• Document delirium subtype (hyperactive, hypoactive, mixed), guides management
• Non-pharmacological measures first (see Section 5)
• Pharmacological intervention if non-pharmacological measures insufficient and patient safety is at risk
• Avoid triggers: remove urinary catheters when safe, minimise physical restraints, treat metabolic derangements

E, Early Mobility and Exercise

• Immobility is a modifiable delirium risk factor. Neuromuscular disuse, loss of proprioceptive inputs, and impaired cerebral autoregulation all contribute
• The ABCDE trial demonstrated feasibility and safety of early mobilisation in mechanically ventilated patients
• A rehabilitation hierarchy: passive range of motion → active range of motion in bed → sitting at edge of bed → standing → walking, progressed according to patient capacity and safety screen (haemodynamic stability, FiO2 ≤ 0.6, RASS ≥ −2, no active resuscitation)
• In ANZ practice, physiotherapists are embedded in most level 3 ICUs and drive the mobilisation program
• Early mobilisation reduces delirium prevalence, duration, and functional disability at discharge

F, Family Engagement and Empowerment

• Family presence reduces disorientation and anxiety. Familiar voices and faces provide re-orientation anchors
• Family-provided reorientation (reminding the patient of time, place, and events) and cognitive stimulation (reading aloud, familiar music, showing family photographs) have face validity and are embedded in most non-pharmacological delirium protocols
• Family communication reduces patient distress and perceived threat, a significant trigger of hyperactive delirium
• In ANZ ICUs, structured family-centred care models (flexible visiting, open visiting hours, family briefings) are standard, enhanced during COVID-era to virtual family presence where necessary
• The ABCDEF bundle's F element also acknowledges family-delivered assessment of baseline cognitive function, critical for identifying pre-existing dementia or cognitive impairment which is the single strongest independent risk factor for ICU delirium

5. Non-Pharmacological Prevention Strategies

Non-pharmacological measures are first-line and should be maximised before pharmacological treatment is considered. The evidence base for individual measures is largely observational, but bundle-based approaches (combining multiple measures) show consistent benefit.

5.1 Sleep Hygiene

Why it matters: Sleep disruption is ubiquitous in the ICU, median polysomnographic sleep efficiency is 50-60%, with severely fragmented architecture, minimal slow-wave sleep, and disrupted circadian rhythms. Sleep deprivation alone can cause delirium in healthy volunteers.

Practical measures:

• Cluster nursing activities to allow 90-minute uninterrupted periods (one full sleep cycle)
• Lights down/off at night; window access for daylight exposure during the day
• Noise reduction protocols: ear plugs, noise-dampening measures, silence at night (telephone volumes, staff conversation)
• Eye masks, shown in RCT to reduce delirium incidence in non-ICU patients; increasingly used in ICU
• Minimise nocturnal monitoring alarms (titrate alarm thresholds to clinical need)
• Melatonin 2-10 mg orally at night for circadian entrainment (low harm, some biological rationale, discussed in Section 6)
• Avoid scheduling non-urgent tests and procedures at night

5.2 Cognitive Stimulation and Reorientation

• Regular verbal reorientation by nursing staff and family members ("You are in hospital, today is Tuesday, the 15th of May...")
• Environmental cues: visible clocks, calendars, family photographs at bedside
• Hearing aids and glasses should be in situ, sensory deprivation is a major predisposing factor
• Cognitive stimulation tasks: familiar music, reading, television news, adjusted to patient tolerance
• Communication aids for non-verbal intubated patients: communication boards, tablet-based apps, alphabet boards

5.3 Mobility (see also E in ABCDEF bundle)

Early mobilisation activates proprioceptive pathways, normalises circadian cortisol rhythms, improves cerebral perfusion, and reduces the inflammatory milieu. Cycle ergometers in bed, in-bed exercises, and progressive out-of-bed activities are all beneficial.

5.4 Optimise Physiology

Address modifiable delirium precipitants:

5.5 Environmental Modification

• Single-room accommodation preferred (reduces noise, allows family presence, reduces hospital-acquired infections)
• Natural lighting cycle preserved where possible
• Minimise unnecessary alarms, overhead PA announcements, and equipment noise

6. Pharmacological Management

6.1 Dexmedetomidine

Mechanism: Selective α-2 adrenergic agonist acting at presynaptic receptors in the locus coeruleus and at spinal dorsal horn neurons. Reduces noradrenergic firing, producing rousable sedation that preserves the thalamocortical network activity associated with natural non-REM sleep. Unlike benzodiazepines, it does not suppress the ARAS via GABA mechanisms.

Pharmacokinetics: Onset 5-10 minutes; elimination half-life 2 hours; hepatically metabolised; minimal renal dosing adjustments required.

Dose in ANZ practice:

• Sedation: 0.2-1.5 mcg/kg/hr IV infusion; usual range 0.5-0.7 mcg/kg/hr
• No bolus loading dose in ICU (causes transient hypertension via peripheral α-2B receptor activation, followed by hypotension)
• Analgosedation for procedural use or NIV: 0.5-1.0 mcg/kg/hr

Evidence for delirium:

• SEDCOM trial (Riker et al. JAMA 2009): Dexmedetomidine vs midazolam in mechanically ventilated patients, dexmedetomidine associated with significantly less delirium (54% vs 76.6%) and shorter time on ventilator. Key contribution: established dexmedetomidine as superior to benzodiazepine sedation for delirium outcomes.

• MENDS trial (Pandharipande et al. JAMA 2007): Dexmedetomidine vs lorazepam, more delirium/coma-free days with dexmedetomidine.

• MENDS2 trial (Hughes et al. NEJM 2021): Dexmedetomidine vs propofol in mechanically ventilated patients with sepsis, no difference in days alive without delirium or coma between groups. Important nuance: propofol is not clearly inferior to dexmedetomidine in this population.

• SPICE-III trial (Shehabi et al. NEJM 2019): Early goal-directed sedation using dexmedetomidine vs standard care in mechanically ventilated patients. No significant difference in 90-day mortality (the primary outcome). Dexmedetomidine was associated with more adverse events (bradycardia, hypotension). Delirium rates were not significantly different. Key take-away: dexmedetomidine does not reduce mortality and may increase haemodynamic adverse events; its superiority in delirium is not unambiguous in an unselected ventilated population.

• For treatment of established delirium (rather than prevention), dexmedetomidine has the strongest evidence among pharmacological agents, particularly for hyperactive delirium in patients who are extubated or on NIV, where it reduces agitation without causing respiratory depression (unlike many other sedatives).

Adverse effects: Bradycardia (most common, dose-dependent; may require dose reduction or atropine), hypotension, rebound hypertension on cessation (taper over 30 minutes), dry mouth.

Practical decision triggers for dexmedetomidine:

• Hyperactive delirium post-extubation where non-pharmacological measures have failed
• Patients where avoidance of respiratory depression is critical (NIV, HFNC)
• Alcohol withdrawal (off-label; useful adjunct, not monotherapy)
• Transition from propofol/midazolam, facilitating SAT and weaning

6.2 Haloperidol

Mechanism: High-potency typical antipsychotic. Dopamine D2 receptor antagonist in the mesolimbic and mesocortical tracts, reduces dopaminergic excess driving perceptual disturbances and psychomotor agitation. Also has mild α-1 adrenergic blockade.

Pharmacokinetics: Oral bioavailability ~60% (first-pass); IV/IM onset 10-20 minutes; half-life ~18-24 hours; hepatic metabolism; active metabolite (reduced haloperidol) has minimal activity.

Dose in ANZ practice:

• Oral: 0.5-2 mg every 6-12 hours, titrated to response; maximum 10 mg/day in most ICU patients
• IV: 0.5-2 mg IV slowly; can repeat every 30-60 minutes for acute agitation (short-term)
• IV haloperidol has lower QTc prolongation risk than IV droperidol but still requires monitoring
• Geriatric/frail patients: start at 0.25-0.5 mg

Evidence:

• MIND-USA trial (Girard et al. NEJM 2018): The most rigorous RCT of antipsychotics for ICU delirium. Haloperidol vs ziprasidone vs placebo in mechanically ventilated patients with hypoactive or mixed delirium. Primary outcome: days alive without delirium or coma (DADC). No difference between any group. No difference in 90-day mortality, ventilator-free days, or safety outcomes. This is the definitive trial.

• HOPE-ICU trial (Page et al. Lancet Respiratory Medicine 2013): IV haloperidol vs placebo in mechanically ventilated patients, no significant difference in DADC at 14 days. Delirium severity was not reduced.

• AID-ICU trial (Andersen-Ranberg et al. NEJM 2022): IV haloperidol 2.5 mg vs placebo in ICU patients with delirium. No significant difference in days alive and out of hospital at 90 days (primary outcome), mortality, or delirium duration. Importantly, no increased harm (including extrapyramidal effects, QTc prolongation).

Synthesis: The weight of RCT evidence does not support routine use of haloperidol for treatment of ICU delirium. It does not shorten delirium duration, reduce days on ventilator, or improve mortality. It remains an option for acute symptomatic management of severely agitated patients when non-pharmacological measures and dexmedetomidine are insufficient, but it should not be initiated with the expectation of treating the underlying delirium.

Adverse effects: QTc prolongation (particularly at higher doses and with IV route, ECG monitoring required), extrapyramidal effects (acute dystonia, akathisia, parkinsonism, more relevant with prolonged use), hypotension (α-1 blockade), neuroleptic malignant syndrome (rare, serious), lowered seizure threshold (relevant in post-operative neurosurgical patients).

Monitoring: 12-lead ECG before initiation; repeat if dose increases; withhold if QTc > 500 ms or increases > 60 ms from baseline.

6.3 Quetiapine

Mechanism: Atypical antipsychotic. D2 antagonism (lower potency than haloperidol), 5-HT2A antagonism, H1 antagonism (responsible for most sedative effect), and α-1 antagonism. The multireceptor profile theoretically addresses multiple neurotransmitter systems implicated in delirium.

Pharmacokinetics: Oral only (no parenteral formulation available in Australia); bioavailability ~9% (extensive first-pass); rapid absorption; half-life ~6 hours; hepatic metabolism.

Dose in ANZ practice:

• 25 mg orally twice daily; titrate to 50-200 mg daily in divided doses
• Maximum 200-300 mg/day in ICU patients (higher doses used in psychiatric practice but not recommended for ICU delirium)
• Only appropriate for patients tolerating enteral medication

Evidence:

• MIND-USA (2018): Ziprasidone (structurally similar atypical antipsychotic, parenteral available) was included rather than quetiapine. No benefit demonstrated.

• Devlin et al. (CCM 2010): Small RCT of quetiapine vs placebo in ICU patients, faster resolution of delirium and shorter time to transfer from ICU in quetiapine group. However, n=36; underpowered; not practice-changing alone.

• No large RCT equivalent to AID-ICU or MIND-USA for quetiapine specifically. Its use is largely extrapolated from pharmacological rationale and small trials.

Practical role: Quetiapine's oral-only formulation limits utility in ventilated patients. It has a role in:

• Post-extubation hyperactive delirium where oral medication is feasible
• Patients with predominant sleep-wake disturbance (H1 antagonism improves sleep quality)
• Patients in whom dexmedetomidine is contraindicated (significant bradycardia, high-grade AV block)

Adverse effects: QTc prolongation (lower risk than haloperidol), postural hypotension, metabolic effects (weight gain, glucose dysregulation, less relevant in short-term ICU use), sedation.

6.4 Benzodiazepines, Avoidance

Benzodiazepines are the most consistently modifiable pharmacological risk factor for ICU delirium. Multiple lines of evidence support their avoidance:

• SEDCOM (2009): Midazolam infusion associated with significantly higher delirium prevalence than dexmedetomidine
• MENDS (2007): Lorazepam infusion independently associated with more delirium/coma-free days lost
• ABCDE Collaborative: Benzodiazepine exposure independently associated with delirium (OR ~1.5-2.5 per day of exposure)
• Mechanistically: non-physiological GABA-A agonism suppresses thalamocortical networks, disrupts sleep architecture, and impairs the acetylcholine and noradrenaline signalling critical for attention

When benzodiazepines are still appropriate in ICU:

• Alcohol withdrawal (benzodiazepines are first-line; CIWA-Ar or fixed-dose protocols)
• Seizure management and status epilepticus
• Procedural sedation (short-acting agents, brief duration)
• Patients with specific indications (e.g. anxiety refractory to other agents, certain drug overdoses, GABA withdrawal states)
• Where other agents have failed in refractory agitation (last resort, acknowledge delirium risk)

Practical rule: If a patient on a benzodiazepine infusion develops delirium, transitioning to propofol or dexmedetomidine should be prioritised.

6.5 Melatonin and Ramelteon

Mechanism: Endogenous melatonin synchronises circadian rhythms via MT1 (sleep onset) and MT2 (phase-shifting) receptors in the suprachiasmatic nucleus. ICU patients have severely blunted nocturnal melatonin peaks.

Evidence: Multiple small RCTs and meta-analyses suggest melatonin (0.5-10 mg nocte) may reduce delirium incidence, but evidence is inconsistent and trials are small. Ramelteon (MT1/MT2 agonist, 8 mg nocte) showed reduction in delirium incidence in a small Japanese RCT (Hatta et al. 2014) but has not been replicated robustly.

ANZ practice: Melatonin is low cost, low harm, and widely available. Most ANZ ICUs use it as part of sleep hygiene protocols (2-5 mg nocte). Ramelteon is not TGA-registered in Australia and is not commonly available.

6.6 Summary Pharmacotherapy Table

7. Special Populations

7.1 Hypoactive Delirium

Hypoactive delirium is the most prognostically serious subtype and the most under-recognised. Patients appear sleepy, withdrawn, and inattentive, nurses may interpret this as the patient "resting comfortably." Only systematic CAM-ICU or ICDSC screening will identify it.

Pharmacological treatment of hypoactive delirium with antipsychotics has even weaker evidence than for hyperactive delirium. The focus should be on non-pharmacological measures, optimising sleep, removing precipitants, and increasing stimulation and mobilisation.

7.2 Alcohol Withdrawal Delirium

Distinct mechanism (GABA withdrawal, glutamate excess) requiring benzodiazepines as first-line. Dexmedetomidine is a useful adjunct, reduces sympathetic tone, agitation, and seizure threshold (though does not prevent seizures; benzodiazepines must be continued). Antipsychotics (haloperidol, quetiapine) may be added for

Sources

• PADIS guidelines (SCCM Pain/Agitation/Delirium/Immobility/Sleep)
• ANZICS guidelines & CORE registry
• Surviving Sepsis Campaign