ICU Delirium: CAM-ICU Screening, Causes and What Actually Works
Delirium is acute brain failure, it is as much an organ dysfunction as a rising creatinine, and the version that kills people is the quiet one nobody charts.
Recognizing delirium, including the kind everyone misses
Delirium is an acute disturbance of attention and awareness that develops over hours to days, fluctuates, and is caused by something: a drug, a metabolic derangement, an infection, the illness itself. The fluctuation is diagnostic gold and also the trap. The patient who was lucid on morning rounds and is climbing out of bed at 22:00 did not "get confused tonight." The delirium was there this morning, cycling through a clear interval while you happened to be watching.
It matters because it is not cosmetic. Delirium in the ICU is consistently associated with longer ventilation, longer stays, higher mortality and long-term cognitive impairment that can look like dementia a year later. Treat it with the seriousness you would give any other acute organ failure, because that is what it is.
Three motor subtypes
- Hyperactive: agitated, pulling at lines, climbing out of bed, hallucinating out loud. The minority subtype, and the one that gets diagnosed every time, because it generates pages and pulls its own arterial line.
- Hypoactive: quiet, withdrawn, sleepy, slow to answer, staring. The majority subtype in the ICU, especially in the elderly, and it carries the worse prognosis.
- Mixed: cycling between the two, often quiet by day and agitated at night.
Hypoactive delirium gets charted as "calm," "resting comfortably," "sleepy after a rough night." Nobody calls you about a quiet patient. But the quiet patient who is inattentive and disoriented has the same brain failure as the screamer, plus a higher risk of aspiration, immobility, and being sedated further because sleepiness was read as comfort. This is exactly why screening exists: hyperactive delirium finds you, hypoactive delirium has to be found. If you only diagnose the delirium that pages you, you are missing most of it.
Two other traps. First, delirium superimposed on dementia: the baseline is abnormal, so the question is never "is she confused" but "is she more confused than her baseline," and the answer lives with the family or the nursing home records, so ask. Second, do not accept "sundowning" as a diagnosis. It is a description of timing. Evening worsening of confusion in a hospitalized patient is delirium until proven otherwise and gets the same workup.
CAM-ICU, step by step
The Confusion Assessment Method for the ICU is a bedside test that takes about two minutes, works in intubated patients because nothing requires speech, and turns "seems okay to me" into a documented, repeatable result. Nursing typically performs it every shift; you should be able to do it yourself, because you will want to repeat it when the story changes.
Step zero: level of arousal (RASS)
Before CAM-ICU you need a Richmond Agitation-Sedation Scale score. RASS runs from +4 (combative) through 0 (alert and calm) to -5 (unarousable). Speak to the patient; if no response, physically stimulate. If the RASS is -4 or -5, the patient is not assessable, and you record exactly that: "unable to assess, RASS -4." A comatose patient is not CAM-ICU negative. At RASS -3 or lighter, proceed.
Feature 1: acute change or fluctuating course
Is the mental status different from baseline, or has it fluctuated in the past 24 hours? Sources: the family, the nurse who had the patient yesterday, the chart. If mental status is at a stable baseline, the screen ends here, negative.
Feature 2: inattention
The core of delirium. The standard test is the letters task: "Squeeze my hand every time I say the letter A," then read a ten-letter sequence such as S, A, V, E, A, H, A, A, R, T at one letter per second. An error is a squeeze on a non-A or a missed squeeze on an A. More than two errors is positive. An intubated patient can do this perfectly well; there is nothing to say. If features 1 and 2 are both present, go on; delirium requires both plus one more.
Feature 3: altered level of consciousness
Simply: is the current RASS anything other than zero? A RASS of -1 or +1 counts. If features 1, 2 and 3 are present, the patient is CAM-ICU positive and you are done.
Feature 4: disorganized thinking
Needed only when feature 3 is absent, meaning the patient is alert and calm but inattentive. Four yes-or-no questions: Will a stone float on water? Are there fish in the sea? Does one pound weigh more than two pounds? Can you use a hammer to pound a nail? Then a command: hold up two fingers, then "now do the same with the other hand" without re-demonstrating. More than one error is positive.
CAM-ICU positive = feature 1 and feature 2, plus feature 3 or feature 4. Chart the RASS and the CAM-ICU result together, every shift, and chart "unable to assess" when that is the truth. A positive screen is not the end of thinking. It is the trigger for the differential below.
The differential: drugs and metabolic causes first
Delirium is a syndrome, not a diagnosis, and the note that says "delirium, continue to monitor" has diagnosed nothing. Something is causing it, the commonest causes are reversible, and the two biggest bins are things you prescribed and things you can measure.
Benzodiazepines are the most consistently implicated drug class in ICU delirium, and a midazolam infusion is the classic offender. Anticholinergics are next: diphenhydramine ordered for sleep or itching, promethazine, oxybutynin, cyclobenzaprine, tricyclics. Opioids contribute, especially meperidine and especially with renal failure stacking metabolites, though untreated pain causes delirium too, so the answer is titration, not abstinence. Add corticosteroids, fluoroquinolones in the susceptible elderly, and the sleep-aid zolpidem. Read the medication administration record yourself, including the PRNs actually given, not just the ordered list, and check doses against the drug guide.
Withdrawal from alcohol, benzodiazepines and opioids, all of which peak on hospital day two to four, exactly when ICU delirium peaks. Ask about home use on admission, because day three is a bad time to learn about the nightly half bottle of wine plus lorazepam. Nicotine withdrawal is real and a patch is cheap. Do not forget the home medications that never got restarted: baclofen and SSRIs have withdrawal syndromes of their own.
Hypoglycemia first, because it is instantly fatal and instantly fixable; check a glucose before you philosophize. Then hyponatremia, hypernatremia, hypercalcemia, uremia, hepatic encephalopathy, hypercapnia and hypoxemia, thyroid disease at the extremes, and thiamine deficiency in the malnourished or alcohol-dependent. Half of this list is on the chemistry panel you already drew; the replacement end of it is covered under electrolyte replacement. A blood gas answers the hypercapnia question that pulse oximetry cannot.
New delirium is an accepted presenting sign of sepsis, and in the frail elderly it may be the only sign, so a fresh screen of cultures, a lactate and a look at the trajectory are reasonable when delirium is new and unexplained. The reverse error also matters: delirium alone is not automatically a urinary tract infection, and reflexively treating bacteriuria ends the workup one step too early.
Stroke, intracranial hemorrhage, nonconvulsive status epilepticus. Head imaging is not routine for every delirium, but it is mandatory for focal deficits, anticoagulation with any new mental status change, head trauma including unwitnessed falls, or delirium with no plausible cause after the survey above. Consider EEG when the patient fluctuates profoundly or twitches, because nonconvulsive status is a great mimic.
Uncontrolled pain, urinary retention, fecal impaction, a full bladder behind a kinked catheter, hypothermia and fever, sleep deprivation from hourly interruptions, and the missing hearing aids and glasses without which a marginal brain cannot orient. Bladder scan, rectal exam, look at the ears, count the overnight wakings. This paragraph solves more delirium than any drug on this page.
Overnight, the sequence is compressed: glucose, vitals with oxygen saturation, a focused neurologic exam, the medication record, a bladder scan, and labs if none are recent. The rest of the overnight approach to acute confusion is on common night calls.
The ABCDEF bundle
The ABCDEF bundle is the closest thing critical care has to a delirium treatment that works, and it is a way of running the whole ICU rather than a drug. Units that perform it well see less delirium, less ventilation and better survival. It is also the rare bundle whose letters are worth memorizing.
Pain first, before sedation, using a scale that works in the nonverbal patient (CPOT or BPS). Untreated pain causes agitation that gets mistreated with sedatives. Analgesia-first sedation, often just fentanyl or hydromorphone with no sedative at all, keeps many ventilated patients comfortable and awake.
Daily, and paired. Turn the sedation off, see who wakes up, and put the ones who wake up on a breathing trial. Paired trials shorten ventilation and get people out of the ICU sooner. The mechanics of the breathing trial side live on the ventilator settings page. The commonest failure is the sedation never actually being paused because the patient "seemed uncomfortable" at 06:00.
Target light sedation, RASS 0 to -1, unless there is a specific indication for deep sedation such as severe ARDS with paralysis or status epilepticus. Prefer propofol or dexmedetomidine over benzodiazepine infusions, which are associated with more delirium and longer ventilation. Every additional point of unnecessary sedation depth is delirium you ordered.
CAM-ICU every shift, hunt the cause when positive, and manage with the nonpharmacologic measures below. This letter is the rest of this page.
The intervention with the best evidence for actually reducing delirium days. Ventilated patients can sit, stand and walk with a good therapy team. Order physical and occupational therapy on day one, not at discharge planning.
Families reorient better than any whiteboard, they know the baseline, and they notice the fluctuation you will miss. Liberal presence at the bedside, and give them a job: bring the glasses, the hearing aids, the photos, the familiar voice.
Nonpharmacologic management comes first
There is no drug that treats delirium. Read that again, because the entire pharmacology section below is downstream of it. What shortens delirium is finding the cause and fixing the environment, and the environmental fixes are cheap, safe and chronically underordered because they do not feel like doctoring.
- Reorient relentlessly. Visible clock, visible date, a window with daylight, staff introducing themselves and saying the day. Simple, repeated, effective.
- Return the senses. Glasses on, hearing aids in and working. A sensory-deprived brain confabulates. This one intervention converts some "agitated and confused" patients into "hard of hearing" patients by lunchtime.
- Protect the night. Lights off, doors closed, alarms tuned down, and consolidate care so the patient is not woken at 24:00, 01:00, 02:00 and 04:00 for tasks that could have been batched. Ask honestly whether the 04:00 labs and the overnight vitals are needed in this patient, and cancel what is not.
- Mobilize by day. Out of bed for meals, therapy as ordered, natural light. Daytime activity is the other half of the sleep protocol.
- Remove the tethers. Every line, catheter and monitor is both a delirium irritant and something to pull out. Ask daily whether the Foley, the telemetry, the restraints and the central line still earn their place. Restraints in particular agitate the agitated; treat them as a bridge measured in hours with a plan to remove, not a standing order.
- Feed, hydrate, bowel and bladder. Dry, impacted, retaining patients are confused patients.
Glasses and hearing aids to bedside. PT and OT consults. Foley out. Telemetry off if not indicated. No overnight vitals if stable. Sleep protocol. Family at bedside with photos from home. Diphenhydramine and zolpidem removed from the PRN list. Melatonin 3 mg at 21:00 is a reasonable, low-risk sleep aid, and while the evidence for prevention is mixed, it displaces worse drugs. None of this is glamorous. All of it works better than what is in the syringe.
When antipsychotics are actually justified
Here is the uncomfortable truth: in randomized trials, antipsychotics have repeatedly failed to shorten delirium, get patients off the ventilator faster, or improve survival, whether the drug was haloperidol or an atypical. They are not a treatment for delirium. They are a chemical restraint with a defined, narrow role, and pretending otherwise is how a night dose becomes a discharge medication.
The narrow role
- Agitation that is dangerous to the patient or staff despite the measures above: actively pulling the endotracheal tube, the fresh arterial line, the ventricular drain.
- Distressing psychotic features: frightening hallucinations or delusions causing real suffering.
- What you are buying is sedation and behavioral control for safety, tonight. You are not treating the underlying brain failure, and the cause-hunt continues in parallel.
The drugs, if you use them
2 to 5 mg IV or IM for acute dangerous agitation, 0.5 to 1 mg in the elderly, repeated at 20 to 30 minute intervals as needed with a low ceiling before you rethink. Check a baseline EKG and monitor the QTc, because IV haloperidol prolongs it and torsades is the complication that ends careers. Avoid it entirely in Parkinson disease and Lewy body dementia.
12.5 to 50 mg PO at night, titrated. The usual choice when parkinsonism forbids haloperidol, and a common scheduled agent when nights are reliably bad. Sedating, orthostasis in the elderly, still a QTc drug.
2.5 to 5 mg PO or dissolving tablet at night. Somewhere between the other two in sedation. Avoid combining the IM form with parenteral benzodiazepines.
An infusion, so an ICU drug, but the right one when agitation is the barrier to extubation or to staying extubated. It sedates without suppressing respiratory drive and is the one sedative with evidence for more ventilator-free time in agitated ventilated patients. Watch for bradycardia and hypotension.
Not for delirium. They cause it. The exceptions are alcohol or benzodiazepine withdrawal, where they are the treatment, and seizures. If lorazepam is being given for "agitation" in a delirious patient outside those settings, stop it.
Write the indication and the stop plan when you write the drug: "quetiapine 25 mg qhs for dangerous nighttime agitation, reassess daily, stop when CAM-ICU negative 24 hours." Antipsychotics started for ICU delirium have a well-documented habit of surviving to discharge and then to the nursing home, in a frail elderly population these drugs carry mortality warnings in. Reconcile them off at every transition. If you started it, you own its funeral.
Documenting delirium
Delirium documentation fails in two directions: the vague note that says "confused, will monitor," and the absent note where hypoactive delirium was never charted at all. Both matter clinically, both matter for coding, and both matter if the chart is ever read in a deposition after a fall or a self-extubation, which is the sort of reading the medical-legal charting page exists for.
Name it, grade it, cause it
- Use the word delirium, with the subtype: "hyperactive delirium," "hypoactive delirium." Not "confused," not "altered," not "sundowning," not "pleasantly confused." For coding, the phrase "acute metabolic encephalopathy" or "acute toxic encephalopathy," with the suspected cause, captures the organ dysfunction in a way "delirium" alone may not, so write both when both are true.
- Chart the objective scores: the RASS and the CAM-ICU result, each shift, including "unable to assess, RASS -4." A dated series of scores is also how you prove later that the delirium was recognized and followed rather than discovered by the incident report.
- Show the cause-hunt: what you checked, what you found, what you changed. "CAM-ICU positive this morning. Reviewed MAR: diphenhydramine given twice overnight, now discontinued. Glucose 112, sodium 134, calcium normal, no hypoxia or hypercapnia on gas. Bladder scan 90 mL. No focal deficit; imaging deferred. Attributed to sedative burden and sleep disruption."
- Justify any antipsychotic with the danger it addressed, the dose, the QTc, and the stop plan, as above.
- Record the baseline and the informant: "Per daughter, sharp at baseline, manages own finances." That sentence is the yardstick every later exam is measured against, and it belongs in the admission history and physical before anyone needs it.
The delirium paragraph in a daily ICU note
Delirium: CAM-ICU positive today (RASS +1, inattentive with 4 errors on letters), hyperactive, day 2. Acute toxic-metabolic encephalopathy, suspected multifactorial: benzodiazepine infusion until yesterday, sleep disruption, hospital day 3 in an 81 year old. Midazolam off, on propofol titrated to RASS 0 to -1, paired awakening and breathing trial this morning, passed SAT. Diphenhydramine removed from PRNs. Glasses and hearing aids at bedside, family in daily, PT and OT following, Foley removed, sleep protocol ordered, melatonin 3 mg qhs. Haloperidol 2 mg IV x1 overnight for pulling at the endotracheal tube; QTc 441 this morning; no scheduled antipsychotic, reassess tonight. Plan: CAM-ICU each shift, expect resolution with sedative clearance; EEG if not improving in 48 hours.
That paragraph names the syndrome, proves it with scores, shows the differential was actually run, documents every intervention including the exit plan for the one drug given, and sets a contingency. It carries forward cleanly into the next day's progress note, where the only thing that should change is the day number and, eventually, the phrase "CAM-ICU negative."
This is not medical advice. It is a teaching outline for clinicians and clinicians in training. Every drug, dose and threshold below is illustrative. Your institution's sedation and delirium protocols, pharmacy and a current reference govern what you actually order.