Sign Up
Login
Try for Free
Login
The CAM-ICU (Confusion Assessment Method for the Intensive Care Unit) is a bedside delirium screen, adapted by Ely and colleagues in 2001 from Inouye's Confusion Assessment Method, that assesses four features once a RASS shows the patient is arousable. ICU and step-down nurses use it every shift to flag delirium for the provider who diagnoses it. This page covers how to document CAM-ICU results, with a fictional sample.
Bedside ICU and step-down nurses at each screen, with intensivists, hospitalists, and advanced practice providers documenting the diagnosis when one is made; no license, training certificate, or publisher qualification level applies, and the rights holder permits implementation and clinical use without written permission
Relieving nurses at handoff, intensivists and hospitalists deciding whether delirium is present, pharmacists reviewing sedatives and other deliriogenic medications, respiratory therapists timing awakening and breathing trials, coders and documentation integrity staff who take the diagnosis from the provider's note, quality staff counting ICU Liberation assessments, and later reviewers
2 to 5 chart lines per screen (RASS with time, features assessed, result and subtype or the reason it could not be done, contributors and action) · the screen itself takes about 2 minutes at the bedside
Clinician-administered delirium screen with a three-way result (positive, negative, or unable to assess) built on four features and gated by a sedation scale; nonverbal responses allow use in intubated patients
Every shift and with any change in mental status in adult ICU patients, before and after spontaneous awakening trials, in step-down units that keep ventilated or recently extubated patients, in trauma and neurological ICUs with the caveats their protocols state, and, in the pCAM-ICU and psCAM-ICU versions, in children's ICUs
Ely and colleagues (Critical Care Medicine and JAMA, 2001) from Inouye's CAM (1990); copyright 2002 E. Wesley Ely and Vanderbilt University, unrestricted for clinical use with the copyright line; named by the SCCM PADIS guideline and the ICU Liberation bundle, mandated by no law; described here for documentation, no worksheet or stimuli reproduced
The Confusion Assessment Method for the Intensive Care Unit is a bedside delirium screen that E. Wesley Ely and colleagues at Vanderbilt adapted from Sharon Inouye's Confusion Assessment Method (Annals of Internal Medicine, 1990) so that it could be used in patients who cannot speak, including patients on a ventilator. Two validation papers appeared in 2001: in Critical Care Medicine, 38 medical and coronary ICU patients and 293 paired daily evaluations, with two study nurses reaching sensitivities of 95 and 96 percent and specificities of 93 percent against a delirium expert applying DSM-IV criteria; and in JAMA, 111 mechanically ventilated patients and 471 paired evaluations, with sensitivities of 100 and 93 percent, specificities of 98 and 100 percent, a kappa of 0.96 between the two nurses, and a screen that took about two minutes. The screen keeps the CAM's four-feature logic. Feature 1 asks whether the patient's mental status has changed acutely from baseline or fluctuated over the past 24 hours. Feature 2 asks whether the patient can attend, tested with a short auditory task the patient answers by hand squeeze, with a visual recognition task as the backup when the auditory task cannot be used. Feature 3 is altered level of consciousness, read straight from the current RASS: any value other than zero counts as present. Feature 4 is disorganized thinking, tested with a few simple yes-or-no questions and a simple command. The screen is positive when Features 1 and 2 are both present and either Feature 3 or Feature 4 is present; when Feature 1 or Feature 2 is absent the screen cannot be positive and the assessment stops, and when Features 1, 2, and 3 are present Feature 4 does not need to be done. Features 3 and 4 swapped numbers when the materials were revised in 2014 (altered consciousness became Feature 3) with no change in content, so older pocket cards and flowsheets label them the other way round. The worksheet, the flowsheet, the letters, pictures, questions, and commands, and the training manual are Vanderbilt's copyrighted materials and are not reproduced here; the CIBS Center's site carries the revised flowsheet and worksheet (the 2023 files), the training manual (updated August 2016), the FAQ (revised March 2014), and translations in more than 20 languages.
The load-bearing fact for documentation is that the CAM-ICU is a screen with three possible results, and only one of them is a delirium finding. Positive means the features of delirium were present at that assessment; negative means they were not; unable to assess means the patient was too deeply sedated or comatose to test content of consciousness, which Vanderbilt's FAQ defines as coma or stupor rather than delirium or normal and restricts to patients at RASS -4 or -5. The RASS therefore comes first at every screen: at the two deepest values the screen is recorded as unable to assess, and most units treat -3 as the deepest value at which it can still be done, some -2. A positive screen is not a diagnosis. Delirium is a medical diagnosis, made and documented by the physician or advanced practice provider from the whole picture (the screen, the sedation and medication history, the examination, and the search for causes), and the ICD-10-CM Official Guidelines assign codes from the provider's documentation, not from a nursing screen. The screen is also weaker at the bedside than in the validation studies: in ten Dutch ICUs, routine nursing CAM-ICUs reached a sensitivity of 47 percent and a specificity of 98 percent against expert assessment (van Eijk and colleagues, 2011), and pooled estimates sit at about 80 to 84 percent sensitivity and 95 to 96 percent specificity, so a negative screen in a patient who looks different to the family or the nurse is a reason to look again, not a reason to stop. Neighbors are charted under their own names: the ICDSC (Bergeron and colleagues, 2001) is an eight-item checklist scored from observations over a period of care rather than at one moment and is the other tool the SCCM guidelines accept; the CAM-ICU-7 (Khan and colleagues, 2017) turns the feature results and the RASS into a 0 to 7 severity score; the pCAM-ICU (Smith and colleagues, 2011) and the psCAM-ICU (Smith and colleagues, 2016) are the pediatric versions; the brief CAM and the Delirium Triage Screen (Han and colleagues, 2013) are the emergency department adaptations; and the Glasgow Coma Scale measures consciousness after brain injury, not delirium.
Critical care nurses chart nearly all of it: a screen every shift under the PADIS guideline's once-per-shift convention, another with any change in mental status, and one before and after each spontaneous awakening trial, each with the RASS that gated it. Intensivists, hospitalists, and advanced practice providers read the results on rounds and write the diagnosis when the screen, the history, and the examination support one; pharmacists read a positive screen against the medication record for benzodiazepines, anticholinergics, and opioid dosing; respiratory therapists time the breathing trial to the awakening trial that makes the screen possible; and quality staff count documented assessments for the ICU Liberation bundle. Step-down nurses use it for ventilated or recently extubated patients who have moved out of the ICU, trauma and neurological ICUs use it with the caveat their protocols state for severe brain injury, and children's ICUs use the pCAM-ICU or psCAM-ICU. The result travels: a transfer-of-care summary should carry the last RASS and CAM-ICU with their times and the next screen due, an incident report after a self-extubation or a fall should record the screen results around the event, a capacity evaluation in the ICU starts with whether delirium is present, and a discharge summary should name an ICU delirium episode so that cognition is followed after discharge. Neighbors win in four places: the ICDSC when the unit has standardized on it, the brief CAM or the 4AT in the emergency department and on general wards where the CAM-ICU was not the validated choice, a cognitive screen such as the Mini-Cog or MoCA once delirium has cleared and the question becomes baseline cognition, and the CIWA-Ar when the question is alcohol withdrawal severity rather than whether delirium is present.
No law, payer, or guideline prescribes a CAM-ICU note format; what exists is a copyrighted screen whose authors recommend documenting it in the hourly nursing flowsheet and, where there is room, at the feature level, a guideline convention of once per shift and as needed, bundle metrics that count documented assessments, and unit protocols that fix the terms and the interval. What survives review is an entry that records the RASS with its time before the screen, names the baseline and its source, records each feature assessed and where the assessment stopped, states the result in the unit's fixed terms with the time and a reason whenever it could not be done, adds the subtype and the contributors reviewed when the screen is positive, records the action and the notification, and leaves the diagnosis to the provider's note while the screen trends toward resolution. Each element below carries the pitfall that most often undermines it.
RASS first, with its time. Open every entry with the RASS value and clock time that gated the screen, and with the stimulus if the value is ambiguous (the RASS page covers that assessment). At -4 or -5 the patient is comatose or too deeply sedated to test content of consciousness: record the CAM-ICU as unable to assess with that reason and set the next attempt, usually after the awakening trial or when the RASS lightens. At -3 and lighter the screen proceeds; Vanderbilt's FAQ notes that most sites use -3 as the lower border and some -2, so name the unit's border in the protocol rather than in each entry. Chart the two values as a pair every time, because Feature 3 is read from that RASS and a reviewer cannot judge a result without it. Pitfall: A CAM-ICU negative at 06:00 beside a RASS of -5 at 06:00, or "UTA" on an awake patient who was simply hard to test.
Baseline mental status and Feature 1. Feature 1 needs a baseline to compare against: write where it came from (family or caregiver, the referring unit, a prior admission, the primary care record) and what it was, including known dementia, hearing or vision loss, language, and a stroke that has left a new permanent baseline. Then state whether there has been an acute change from that baseline or fluctuation over the past 24 hours, drawing on the night entries, handoff, and the family's account; a family report of confusion the nurse did not see counts and is charted as collateral. Feature 1 evaluates the day and Feature 3 the moment: a patient who fluctuated overnight and is now alert at RASS 0 is Feature 1 present and Feature 3 absent, and the entry should read that way. Pitfall: No baseline anywhere, so a patient with dementia is charted as acutely changed, or a night of fluctuation is erased by a calm 08:00 assessment.
The features assessed, and where the assessment stopped. Record each feature assessed as present or absent, and say which were not assessed because the rule made them unnecessary: if Feature 1 or Feature 2 is absent the screen stops and is negative; if Features 1, 2, and 3 are present the screen is positive without Feature 4; Feature 4 is needed only when Features 1 and 2 are present and the patient is at RASS 0. For Feature 2 note which attention task was used and, if the visual backup was needed, why (hearing loss, language); for Feature 4 note when the questions alone were scored because the patient could not move an arm. Most flowsheets store only the overall result; Vanderbilt's FAQ says feature-level documentation helps compliance and accuracy and gives chart review something to audit, and the note of which features were tested is what makes a positive reproducible. Pitfall: A bare "CAM+" with no feature results, or all four features charted as tested on a patient whose screen should have stopped at Feature 2 absent.
Result, time, subtype, and the reason when it could not be done. Write the result in the unit's fixed vocabulary (positive or negative, yes or no, present or absent, with unable to assess as the third value) and the clock time, and do not mix vocabularies across shifts. Unable to assess always carries its reason: RASS -4 or -5 with the sedation or the coma named, pharmacologic paralysis, or another stated barrier; it is never a synonym for negative and never a way to record a difficult patient. When positive, add the motoric subtype by the RASS the way the unit protocol defines it, usually hypoactive at a negative or zero RASS and hyperactive or mixed at a positive one, because hypoactive delirium is the common form and the one most often missed: in 614 medical ICU patients, purely hyperactive delirium was 1.6 percent of delirium, hypoactive 43.5 percent, and mixed 54.9 percent (Peterson and colleagues, 2006). Chart the prior result and its time beside the new one so the fluctuation is visible. Pitfall: UTA with no reason, or a positive with no subtype, so three quiet days of hypoactive delirium are never named.
Contributors reviewed. A positive screen is a prompt to look for causes, and the entry should show that the look happened: sedatives and benzodiazepines on board or given overnight, anticholinergics and other deliriogenic drugs, opioid dosing against the pain score and its tool, sleep and the night's interruptions, oxygenation, infection markers and the team's workup, electrolytes and glucose, restraints, catheters and lines, and whether the hearing aids and glasses are in. Name the pharmacist's review when one was made, and the alcohol or sedative withdrawal question when the history raises it (the CIWA-Ar owns that assessment). The list is the nursing half of a differential the provider completes; it is not a diagnosis. Pitfall: A positive screen followed by "delirium precautions" with nothing reviewed, or a contributor list that repeats unchanged for four shifts.
Action, notification, and the family. Record what was started or changed: reorientation, sleep protection, mobility with the therapist, devices removed, family presence, the medication changes made under orders, and the awakening and breathing trials that day, which are the bundle elements a positive screen is meant to trigger. Record who was notified, at what time, and what they decided, including a decision to make no drug change. No guideline requires a medication in response to a positive screen: the SCCM PADIS guideline (2018) suggests against routine antipsychotics for treatment, and the 2025 focused update was unable to recommend for or against them, so any antipsychotic given is charted against a specific indication such as distressing or dangerous agitation, with the order and the response, rather than against the screen result. Note the family conversation and the information given. Pitfall: A positive screen with no action and no notification time, or an antipsychotic charted as the response to a positive CAM-ICU with no indication stated.
Diagnosis, trend, and handoff. Keep the two documents in their lanes. The nursing entry reports the screen; the provider's progress note states the diagnosis (delirium, its likely contributors, and the plan) and is the note a coder can use, because the ICD-10-CM Official Guidelines base code assignment on the provider's documentation and list no delirium screen among the exceptions that allow another clinician's note. Chart the results serially with times so the course is visible, repeat the screen when a UTA patient lightens, and, if the unit follows Vanderbilt's convention that a patient is free of delirium after 24 hours of negative screens, say so with the times. At handoff and transfer carry the last RASS and CAM-ICU with their times, the subtype, the contributors still open, the medication changes, and the next screen due; the transfer-of-care summary is where that line lives. Pitfall: A nursing note that reads "patient is delirious" as a diagnosis, or a handoff of "CAM negative" with no time, no RASS, and no next screen due.
CAM-ICU DOCUMENTATION BLOCK (screen entry with gate and context) Date: [ ] Time: [ ] Setting: [ICU / step-down / trauma or neuro ICU / PICU] Nurse: [ ] Version: [CAM-ICU / pCAM-ICU / psCAM-ICU] Unit terms: [positive, negative, UTA / yes, no, UTA] RASS at the screen: [ ] Time: [ ] (if -4 or -5: CAM-ICU unable to assess, see below) Baseline mental status: [ ] Source: [family / referring unit / prior record / primary care] Known dementia, hearing or vision loss, language, prior stroke: [ ] Feature 1, acute change or fluctuation over 24 hours: [present / absent] Evidence: [night entries, handoff, family account] Feature 2, inattention: [present / absent] Task used: [auditory / visual backup, reason] Feature 3, altered level of consciousness (current RASS other than 0): [present / absent] Feature 4, disorganized thinking: [present / absent / not needed] [questions alone: reason] Result: [positive / negative / unable to assess] Time: [ ] Unable to assess, reason: [RASS -4 or -5 with sedation or coma named / paralysis / other] Subtype if positive (by RASS, per unit protocol): [hypoactive / hyperactive / mixed] Prior result and time: [ ] Pattern over 24 hours: [ ] Contributors reviewed: [sedatives, benzodiazepines / anticholinergics / opioids and pain score with tool / sleep / oxygenation / infection / electrolytes, glucose / restraints, catheters, lines / hearing aids, glasses] Pharmacist review: [who, time, findings] Action: [reorientation / sleep protection / mobility / devices removed / family presence / medication changes under orders / awakening and breathing trial today] Notification: [provider, time, decision] Family: [present, informed, baseline confirmed] Provider diagnosis: [documented in the provider's note / not established] (the screen is not a diagnosis) Next screen due: [ ] Handoff line: [last RASS and CAM-ICU with times, subtype, open contributors, medication changes, next due] Nurse signature / credentials: Date:
Free to use and share, no signup. The PDF includes a one-page cheat sheet with element-by-element pitfalls and a pre-sign checklist; the DOCX is the blank documentation block, ready to adapt. Neither reproduces the worksheet, flowsheet, stimuli, or training manual.
Scenario: a US community teaching hospital's surgical ICU, day two after an emergency laparotomy in a 72-year-old man who is still ventilated; the day nurse charts an unable-to-assess entry at deep sedation, a positive screen after the awakening trial with the features assessed in sequence, the contributors reviewed with the pharmacist, the notification, the intensivist's separate diagnosis, an evening reassessment, and the run of negative screens that follows. All details are fictional.
Patient: H.B., 72 · Setting: Surgical ICU, community teaching hospital; day 2 after emergency laparotomy · Clinician: M. Okafor, RN, CCRN · Note date: 09/16/2026
Baseline, gate, and an unable-to-assess entry (06:10): CAM-ICU as built in the unit flowsheet (results recorded as positive, negative, or UTA), assessed by M. Okafor, RN, on 09/16/2026. Baseline from his daughter by phone at 06:30 and from the admission history: independent at home, managing his own medications and finances, no diagnosis of dementia, wears hearing aids in both ears (at home overnight, brought in this morning), reading glasses. Night entries: CAM-ICU negative at RASS 0 at 20:15 on 09/15/2026; RASS -1 to -2 after midnight; RASS +2 at 02:10 while pulling at his gastric tube, then RASS -3 at 02:40 after a propofol bolus, with no screen attempted between; propofol increased at 03:30 for ventilator dyssynchrony. 06:10: RASS -4, no eye opening to his name, movement to physical stimulation only. CAM-ICU: unable to assess, reason deep sedation on propofol 40 mcg/kg/min; not recorded as negative. CPOT 0. Next attempt set for after the spontaneous awakening trial planned with the respiratory therapist (K. Brandt, RRT) at 08:30, under the unit protocol (SICU-DEL-02), which sets a screen every shift, with any change in mental status, and after each awakening trial.
Awakening trial, the gate passed, and the screen (09:40): Propofol stopped at 08:30 for the awakening trial; fentanyl 50 mcg/h continued. 09:35: RASS -1, opens his eyes to his name and holds contact briefly, then drifts; no physical stimulation needed, so the screen proceeds. Hearing aids in place from 09:00. Feature 1 present: acute change from an independent, oriented baseline, and fluctuation over the past 24 hours (the 02:10 agitation, the sedated interval, the current drowsiness). Feature 2 present: inattention on the auditory attention task by hand squeeze; the visual backup was not needed with the hearing aids in. Feature 3 present: current RASS -1. Feature 4 not assessed: not needed once Features 1, 2, and 3 were present. Result at 09:40: CAM-ICU positive. Subtype recorded as hypoactive under the unit protocol (positive screen at a negative RASS, quiet, little spontaneous movement). Prior result: UTA at 06:10; last completed screen before that: negative at RASS 0 at 20:15 on 09/15/2026.
Contributors reviewed and actions (09:45 to 11:30): Pain: CPOT 3 with grimacing on turning at 09:45, fentanyl 25 mcg bolus under the pain order, CPOT 1 at 10:15. Sedation: propofol off since 08:30 and not restarted (RASS -1 to 0 through the morning against the ordered target of -1 to 0 written 09/15/2026); midazolam 2 mg intravenous given three times overnight for agitation per the medication administration record; diphenhydramine 25 mg on the bedtime order given at 22:30. Pharmacist review (R. Delgado, PharmD) at 10:20: midazolam and diphenhydramine flagged as deliriogenic, with a recommendation to the intensivist to discontinue both and to treat agitation with pain reassessment first. Sleep: interrupted every hour overnight per the flowsheet. Oxygenation: SpO2 94 to 97 percent on FiO2 0.40. Infection: afebrile, white cell count and cultures reviewed by the team on rounds; sodium 133 and glucose 162 noted for the team. Devices: urinary catheter day 2, reviewed for removal; two peripheral lines. Actions: hearing aids and glasses in; whiteboard updated with the date, the place, and the daughter's visit time; blinds opened; reorientation at each contact; physical therapy sat him at the edge of the bed at 11:10; daughter at the bedside from 11:00 and told what a positive screen means and does not mean. Spontaneous breathing trial started 09:50 by K. Brandt, RRT, and passed at 10:20; extubation decision deferred to rounds.
Notification and the provider's diagnosis: Intensivist (Dr. S. Nakamura) notified at 09:50 of the positive screen, the RASS, the features, and the pharmacist's findings; at 10:40 she discontinued midazolam and diphenhydramine, ordered melatonin at bedtime, kept the sedation target at -1 to 0 with no continuous sedative, and asked for the screen to be repeated on the evening shift. Her progress note for 09/16/2026 documents the diagnosis, hypoactive delirium, postoperative and multifactorial, with sedative exposure, sleep disruption, and pain named as contributors, and the plan above. This nursing entry reports the screen and the actions; the diagnosis, and the code that follows from it (F05), come from her note and not from the flowsheet. No antipsychotic was ordered or requested; the intensivist recorded that the indication for one was absent.
Evening reassessment (20:30): Extubated at 15:40 after rounds, on nasal cannula. RASS 0 at 20:30, awake and calm, no sedative running. Feature 1 present: fluctuation over the past 24 hours. Feature 2 present: inattention on the auditory task, hearing aids in. Feature 3 absent: RASS 0. Feature 4 assessed because Feature 3 was absent: present, with disorganized answers to the questions and the command not followed. Result at 20:30: CAM-ICU positive, hypoactive by the protocol's RASS rule (RASS 0, quiet). CPOT 1; melatonin given at 21:00; lights dimmed and care clustered for the night; daughter reoriented him before leaving at 21:30. Night nurse to screen again with any change in behavior and at the start of the next shift.
Resolution and handoff (09/17/2026 to 09/18/2026): 09/17/2026 08:15: RASS 0; Feature 1 present (fluctuation within the past 24 hours); Feature 2 absent, so the screen stopped: CAM-ICU negative. 20:20: RASS 0; Feature 1 present; Feature 2 absent: negative. 09/18/2026 08:05: RASS 0; Feature 1 absent (no change or fluctuation in the past 24 hours); screen stopped: negative, and 24 hours of negative screens recorded, the unit's convention for delirium resolved, with the intensivist documenting resolution in the 09/18/2026 progress note. Screening continues every shift while he remains in the unit. Handoff line at 18:45 on 09/18/2026: RASS 0 at 18:30; CAM-ICU negative at 08:05, next due 20:00; hypoactive delirium 09/16/2026 to 09/17/2026, resolved per the provider's note; midazolam and diphenhydramine discontinued, melatonin nightly; hearing aids and glasses at the bedside; daughter visits at 11:00 and 19:00. Every result in this record sits beside the RASS and the time that gated it, and no result was entered as negative when the screen could not be done.
This sample is fictional and for educational purposes. It does not describe a real patient or record; the values, times, doses, dates, and details are invented to show documentation structure and are not clinical guidance. No worksheet, stimuli, or training manual text is reproduced.
Writing these after every session? BastionGPT drafts complete notes from bullets, dictation, or a transcript.
Generate a note from bulletsUnited States: no LAW names the CAM-ICU or sets a delirium screening interval, and the rules that bind sit one step from the screen. Under LAW, the Medicare Conditions of Participation require a medical record with nursing notes, medication records, and the information needed to monitor the patient's condition, with entries complete, dated, timed, and authenticated (42 CFR 482.24), and a nursing assessment and care plan for every inpatient (42 CFR 482.23); a CAM-ICU entry meets them when a reader can see what was assessed, at what RASS, at what time, with what result and response. The FY 2026 ICD-10-CM Official Guidelines (Section I.B.14) base code assignment on the documentation of the patient's provider and list the exceptions where another clinician's note may be used (body mass index, pressure ulcer stage, coma scale, the NIH stroke scale, and a few others); a delirium screen is not among them, so the diagnosis behind F05 has to be in the physician's or advanced practice provider's note, and a nursing flowsheet reading positive does not create it. Under PAYER POLICY, CMS's Age Friendly Hospital Measure, adopted in the FY 2025 inpatient prospective payment rule for the Hospital IQR Program beginning with the calendar year 2025 reporting period and the FY 2027 payment determination, is a structural attestation measure with five domains; its Frailty Screening and Intervention domain asks hospitals to attest that they screen patients 65 and older for mentation (delirium and cognitive impairment), mobility, and malnutrition with validated instruments and act on positive screens, and the first attestations were due through the Hospital Quality Reporting system by May 15, 2026 (CMS specifications and attestation guide, July 2025). It names no instrument and sets no interval; failure to report carries the IQR consequence and the attestation is the deliverable, so a unit's CAM-ICU process is evidence for the attestation rather than a measure in its own right, and no CMS national coverage rule, NCCI edit, or Hospital IQR outcome measure keys payment to a CAM-ICU result. The American College of Surgeons Geriatric Surgery Verification program carries delirium screening for older surgical patients as a program standard (CONVENTION with verification force); a June 2026 study in the Journal of the American College of Surgeons found 94.3 percent of older surgical patients screened at verified hospitals against 52.5 percent elsewhere, with similar positive rates, which is the measurable cost of not screening. Under CONVENTION, the SCCM PADIS guideline (Devlin and colleagues, 2018) states as a good practice statement that critically ill adults should be regularly assessed for delirium using a valid tool, names no frequency, and suggests against routine antipsychotics for prevention and for treatment (conditional, low quality evidence); Vanderbilt's FAQ carries the guideline convention as every shift (every 8 to 12 hours) and as needed, with a screen before and after each awakening trial because Patel and colleagues (2014) found that delirium which resolved within two hours of stopping sedatives, 12 percent of 102 patients, carried a prognosis like no delirium while persistent delirium did not. The 2025 focused update (Lewis and colleagues) was unable to recommend for or against antipsychotics for treatment, and the MIND-USA trial behind that shift randomized 566 delirious patients across 16 centers, 89 percent of them hypoactive, and found median days alive without delirium or coma of 7.9 with haloperidol, 8.7 with ziprasidone, and 8.5 with placebo (P 0.26) (Girard and colleagues, 2018). The SCCM ICU Liberation data collection manual (2025) turns the convention into a metric: D element compliance is at least two documented delirium assessments per 24 hours with the CAM-ICU or ICDSC, a finding of unable to assess is considered aligned with the goal, and patients in coma or at RASS -4 or -5 are excluded, which is the definition a quality team should cite before deciding whether a UTA counts as screened, because other programs define it differently. Vanderbilt's own trauma delirium guideline (revised June 2026) states that delirium is a clinical diagnosis made by the team from the injuries, GCS, RASS, and CAM-ICU together, that the CAM-ICU cannot be assessed at RASS -4 or -5 and should only be reported as unable to assess, and that in severe traumatic brain injury the screen is not solely diagnostic. The Joint Commission has no CAM-ICU-specific standard. Three habits deserve the same label: a once-daily screen as the standard, a sedated patient charted as negative, and an antipsychotic as the automatic response to a positive screen are unit habits, and no authority above states any of them.
Canada and Australia name a tool only by example and set no interval, and the version and accuracy boundaries hold everywhere. In Canada, no federal or provincial LAW names the CAM-ICU; the Canadian Coalition for Seniors' Mental Health national guidelines on the assessment and treatment of delirium (2006, with a 2014 update, and a 2010 guideline for delirium at the end of life) are CONVENTION for older adults, and no delirium-specific Accreditation Canada Required Organizational Practice was located for this page. The clearest critical care guidance is provincial: Critical Care BC's ICU Liberation bundle (Provincial Health Services Authority, 2025; CONVENTION) states that delirium should be routinely monitored in all ICU patients with a validated tool, the CAM-ICU or the ICDSC, at least once a shift and more often with any change in mental status, that delirium is reported on daily interdisciplinary rounds with the medications and environmental factors that may be contributing, that the first step in a positive patient is to identify the cause before any drug, and that no data support routine antipsychotics for prevention. Health Standards Organization's HSO 11001:2026 Critical Care Services was in public review as a draft dated March 2026 as of September 2026 and asks for validated assessment tools with reassessment intervals defined by the organization, naming no delirium tool, so it should be cited as a draft. In Australia the frame is firmer. The ACSQHC Delirium Clinical Care Standard (2016, revised September 2021, with assessments against the revised standard from 1 September 2022; CONVENTION with accreditation force) carries eight quality statements, asks for a validated tool rather than naming one, and states that routine antipsychotic use is not recommended; action 5.29 of the NSQHS Comprehensive Care Standard (an accreditation requirement for Australian hospitals) requires routine screening for cognitive impairment with a validated tool for patients 65 and over, for patients of any age at risk of delirium, and when family or carers raise concerns, with results documented and communicated, and states that a positive score on a screening tool is a prompt for further assessment, not a diagnosis. Delirium is hospital-acquired complication 11 on the national list, with a 2019 to 2020 rate of 35.7 per 10,000 admissions quoted by the Commission; IHACPA has applied a funding reduction to episodes with a hospital-acquired complication since July 2018 (PAYER POLICY), and its 2026 to 2027 pricing documents state that under clause A90 of the Addendum to the National Health Reform Agreement 2026 to 2031 the complication penalties are shadow priced rather than applied while the Commission reviews the safety and quality measures, so an Australian ICU should describe the current year as shadow priced rather than penalized. Three boundaries apply in every country. Version: the pCAM-ICU (Smith and colleagues, 2011) was validated in 68 children aged at least 5 years, chronologically and developmentally, across 146 paired assessments (sensitivity 83 percent, specificity 99 percent, kappa 0.96), and the psCAM-ICU (Smith and colleagues, 2016) in 300 children aged 6 months to 5 years across 530 paired assessments (sensitivity 75 percent, specificity 91 percent, kappa 0.79), so a children's ICU names the version and does not import adult conventions; the CAM-ICU-7 is a severity score, not the screen; and the brief CAM and Delirium Triage Screen belong to the emergency department (Han and colleagues, 2013: 406 patients, the triage screen 98 percent sensitive, the brief CAM 84 percent sensitive and 96 percent specific in physicians' hands). Accuracy: research nurses reached sensitivities of 93 to 100 percent in the 2001 studies, pooled estimates are 80.0 percent sensitivity and 95.9 percent specificity across nine studies and 969 patients (Gusmao-Flores and colleagues, 2012) and 0.84 and 0.95 across 29 studies (Chen and colleagues, 2021, who found accuracy shaped by the share of hypoactive delirium, ICU type, and mechanical ventilation), and routine nursing screens in ten Dutch ICUs reached 47 percent sensitivity with 98 percent specificity (van Eijk and colleagues, 2011), so a negative screen rules out less than the validation papers suggest, and the ICDSC (Bergeron and colleagues, 2001; eight items, predicted sensitivity 99 percent and specificity 64 percent in the original study) trades in the opposite direction. Population: the FAQ states the screen is valid with and without dementia but harder, and asks the family whether the patient could do the tasks at baseline; it was validated in 129 post-stroke patients (sensitivity 76 percent, specificity 98 percent) and the authors ask neurological ICUs to read a positive as symptoms of delirium rather than definitely delirium; a blind patient or one who cannot move an arm is scored on the questions alone for Feature 4; and a language barrier is met with one of the more than 20 translations or charted as the reason the screen could not be completed, never as negative.
The CAM-ICU and its educational materials are copyrighted: the copyright line is Copyright © 2002, E. Wesley Ely, MD, MPH and Vanderbilt University, all rights reserved, and it appears on the revised flowsheet and worksheet as well as the 2016 training manual and the 2014 FAQ. Vanderbilt's Critical Illness, Brain Dysfunction, and Survivorship Center states that it has deliberately made the materials unrestricted in terms of use, asks that the copyright line appear at the bottom of pocket cards and other educational materials, does not require a written letter of permission for implementation and clinical use, and asks anyone who wants to use the materials for other uses to contact [email protected] for permission; as of September 2026 its public pages publish no separate terms for EHR builds, publications, or commercial products, so those are the other uses that go to that address rather than rights this page can describe. The RASS printed on the CAM-ICU pocket card is Virginia Commonwealth University's copyrighted scale under its own terms, and the original CAM belongs to Dr. Inouye and the Hospital Elder Life Program. BastionGPT is not affiliated with, or endorsed by, the authors or the rights holder. This page reproduces no test items, stimuli, norms, or scoring materials.
The numbers behind these errors are specific. In the validation studies, two research nurses reached sensitivities of 100 and 93 percent with specificities of 98 and 100 percent across 471 paired evaluations in 111 ventilated patients (Ely and colleagues, JAMA 2001); in routine practice across ten Dutch ICUs, nursing screens reached a sensitivity of 47 percent with a specificity of 98 percent after 101 of 282 patients were excluded as comatose (van Eijk and colleagues, 2011), and pooled estimates sit at 80.0 percent sensitivity and 95.9 percent specificity (Gusmao-Flores and colleagues, 2012) and 0.84 and 0.95 (Chen and colleagues, 2021). Implementation studies show that completion and accuracy are different problems: at two medical centers, compliance with the CAM-ICU was 90 and 84 percent across 711 patients and 4,163 patient-days, with kappas of 0.92 and 0.75 against reference raters (Pun and colleagues, 2005); in a trauma ICU, nurses completed it in 84 percent of evaluations with a kappa of 0.77 overall and 0.62 in ventilated patients against 1,011 expert assessments (Soja and colleagues, 2008); and in a surgical ICU, a one-week nursing education campaign cut inappropriate unable-to-assess ratings from 32 to 19 percent, with ventilated patients far more likely to receive one (Swan, 2014). In 614 medical ICU patients, purely hyperactive delirium made up 1.6 percent of delirium, hypoactive 43.5 percent, and mixed 54.9 percent (Peterson and colleagues, 2006), which is why an unnamed subtype is usually a missed hypoactive case. No published audit counts how often a CAM-ICU entry lacks its RASS, its time, or its feature results; the bundle metrics and coding rules described under compliance considerations are what turn those omissions into a quality or coding finding. The BastionGPT Clinical Advisory Board sees the same errors most often in CAM-ICU documentation reviews:
BastionGPT is specifically trained, tuned, and clinically tested on critical care nursing and delirium screening documentation.
See how clinicians use it day to day on the AI medical notes page.
Many BastionGPT users report saving more than 90 minutes per day on documentation.
HIPAA-compliant with a signed BAA on every plan. Your data is never used to train models. BastionGPT drafts, you review and sign.
The CAM-ICU is not added up; it is a decision built from four features once the patient is arousable. Feature 1 asks whether mental status has changed acutely from baseline or fluctuated over the past 24 hours. Feature 2 asks whether the patient can pay attention, tested with a short auditory task the patient answers by squeezing the assessor's hand, with a visual recognition task as the backup for patients who cannot hear the task. Feature 3 is altered level of consciousness and is read straight from the current RASS: any value other than zero counts as present. Feature 4 is disorganized thinking, tested with a few simple yes-or-no questions and a simple command. The screen is positive when Features 1 and 2 are both present and either Feature 3 or Feature 4 is present. Only the features needed to reach the answer have to be done: if Feature 1 or Feature 2 is absent the screen is negative and stops, and if Features 1, 2, and 3 are present the screen is positive without Feature 4, which is needed only when the patient is at RASS 0. The letters, pictures, questions, and commands, and the error thresholds for Features 2 and 4, are on Vanderbilt's copyrighted worksheet and flowsheet and are not reproduced here; train from the current materials on icudelirium.org, because Features 3 and 4 swapped numbers in the 2014 revision and older cards label them the other way round.
Chart the RASS with its time first, every time, and then the screen. At RASS -4 or -5 the patient is comatose or too deeply sedated to test content of consciousness, and Vanderbilt's FAQ is explicit that the result is unable to assess, which means coma or stupor rather than delirium or normal, and that UTA should be used only for comatose patients; the SCCM ICU Liberation manual and Vanderbilt's own trauma guideline treat -4 and -5 the same way. At -3 the FAQ says most patients can still be rated and most sites use -3 as the lower border, with some choosing -2, so the border is a unit protocol decision that belongs in the protocol, not in each entry. Any lighter value clears the gate, including agitated values. Write UTA with its reason (RASS -4 on propofol, coma, pharmacologic paralysis) and the next attempt, usually after the awakening trial; never as a synonym for negative, and never for an awake patient who is slow, deaf, intubated, or speaks another language, because those are testing problems with solutions, not coma. A one-week education campaign in one surgical ICU cut inappropriate UTA ratings from 32 to 19 percent, and ventilated patients were the ones most often mislabeled (Swan, 2014). Whether a UTA counts as a completed screen for your quality report depends on the program: the ICU Liberation manual considers a UTA finding aligned with its goal and excludes patients in coma or at RASS -4 or -5, while other programs define it differently, so cite the definition you use. The RASS page covers the assessment that produces the gating value.
No law or accreditor in the United States, Canada, or Australia sets a CAM-ICU interval, and neither does PADIS: the 2018 guideline states as good practice that critically ill adults should be regularly assessed for delirium using a valid tool and leaves the frequency to the unit. The working convention, carried in Vanderbilt's FAQ from the 2013 PAD guideline, is every shift (every 8 to 12 hours) and as needed, with more frequent screens when the patient's condition changes; Vanderbilt also recommends screening before and after each spontaneous awakening trial, because delirium that clears within two hours of stopping sedatives carries a different prognosis from delirium that persists (Patel and colleagues, 2014). Critical Care BC says at least once a shift and more often with any change in mental status; Vanderbilt's trauma guideline says each shift with the results reported on rounds; the SCCM ICU Liberation data manual counts at least two documented assessments per 24 hours as compliant. Once-daily screening is a habit, not a standard, and it misses fluctuation by design. Whatever interval your protocol sets, cite the protocol in the record as policy, chart the RASS with each screen, and add an unscheduled screen whenever behavior changes, because the value of the tool is serial.
No. A positive screen means the features of delirium were present at that assessment; delirium itself is a medical diagnosis, made by the physician or advanced practice provider from the screen, the history, the examination, the medication record, and the search for causes, and documented in the provider's note. Vanderbilt's trauma guideline puts it plainly: delirium is a clinical diagnosis made by the team from the injuries, GCS, RASS, and CAM-ICU together, and in severe brain injury the screen is not solely diagnostic; the FAQ asks neurological ICUs to read a positive as symptoms of delirium rather than definitely delirium. The coding rules follow the same line: the ICD-10-CM Official Guidelines base code assignment on the provider's documentation and list the exceptions where another clinician's note may be used (body mass index, pressure ulcer stage, coma scale, NIH stroke scale, and a few more), and a delirium screen is not one of them, so F05 comes from the provider's note and never from the flowsheet. The nursing entry therefore reports the screen result, the features, the contributors reviewed, the actions, and the notification, and says that the provider was informed; it does not read "patient is delirious." If you believe the patient is delirious and the provider disagrees, chart your objective findings and the conversation, repeat the screen, and ask for the reassessment; a screen result recorded as observed is not a claim about the diagnosis. Where the diagnosis is made, the discharge summary should carry it so that cognition is followed after the ICU.
Intubation is never a reason to skip it: the screen was built for ventilated patients, the attention task is answered by hand squeeze, the questions by nodding or squeezing, and the command is nonverbal, so chart the screen as usual and note intubated for context. Under neuromuscular blockade the patient cannot respond at all, so chart unable to assess with the reason (pharmacologic paralysis, the agent, and its start time), carry the last result before the blocker with its time, and screen again when the block has worn off and the RASS is -3 or lighter; the same logic applies to the RASS under paralysis, which the RASS page covers. A patient who is blind or cannot move an arm is scored on the Feature 4 questions alone, per Vanderbilt's FAQ, and a patient who cannot hear the auditory task uses the visual backup; write which task was used and why. For a language barrier, use one of the more than 20 translations of the worksheet if the unit has it, or an interpreter for the questions, and if neither is possible chart that the screen could not be completed and the reason, never negative. In every one of these cases the entry names the barrier, what was done about it, and when the next attempt is due.
Four things, in the order the bundle expects. Subtype: name it by the RASS the way your protocol defines it, usually hypoactive at a negative or zero RASS and hyperactive or mixed at a positive one; hypoactive delirium is the common form and the one most often missed, and in 614 medical ICU patients purely hyperactive delirium was 1.6 percent of delirium (Peterson and colleagues, 2006). Contributors: list what you reviewed, named sedatives and benzodiazepines, anticholinergics, opioid dosing against a pain score with its tool, sleep and its interruptions, oxygenation, infection markers, electrolytes and glucose, restraints, catheters and lines, and whether hearing aids and glasses are in, and note the pharmacist's review when made; withdrawal is assessed under its own tool (the CIWA-Ar for alcohol). Action: reorientation, sleep protection, mobility, devices removed, family presence, the medication changes made under orders, and the awakening and breathing trials that day, with the provider notified at a stated time and their decision recorded. Medication: no guideline requires a drug for a positive screen; PADIS 2018 suggests against routine antipsychotics for prevention and for treatment, the 2025 focused update was unable to recommend for or against them for treatment, and MIND-USA found no difference in days alive without delirium or coma among haloperidol, ziprasidone, and placebo in 566 patients (Girard and colleagues, 2018), so an antipsychotic, when given, is charted against a specific indication such as distressing or dangerous agitation with the order and the response, never against the screen result. Then screen again at the interval the protocol sets and whenever behavior changes.
Use the version validated for the setting and name it. The adult CAM-ICU was validated in ICU patients, including ventilated ones, and has emergency department validation (Han and colleagues, Academic Emergency Medicine, 2014); a step-down unit that keeps ventilated or recently extubated patients can use it with the same gate and terms, and Vanderbilt supports delirium assessment throughout the hospital while recommending setting-appropriate tools. In the emergency department the two-step Delirium Triage Screen and brief CAM (Han and colleagues, 2013: 406 older patients, the triage screen 98 percent sensitive as a rule-out, the brief CAM 84 percent sensitive and 96 percent specific in physicians' hands) were built from the CAM-ICU for that setting; on general wards the CAM, the 3D-CAM, the 4AT, and the Nursing Delirium Screening Scale are the usual choices, and Australian hospitals meeting action 5.29 use whichever validated tool the service has adopted. For children, the pCAM-ICU covers children at least 5 years old both chronologically and developmentally (Smith and colleagues, 2011) and the psCAM-ICU covers infants and children from 6 months to 5 years (Smith and colleagues, 2016), so the choice between them follows developmental as well as chronological age; the SCCM ICU Liberation manual also accepts the Cornell Assessment of Pediatric Delirium. Whatever you use, the same documentation rules hold: the arousal check first, the features assessed, the result with its time and reason, and the diagnosis left to the provider.
For your own clinical use, yes, on Vanderbilt's stated terms. The CAM-ICU and its educational materials are copyrighted (Copyright 2002, E. Wesley Ely, MD, MPH and Vanderbilt University, all rights reserved), and the Critical Illness, Brain Dysfunction, and Survivorship Center states that it has deliberately made them unrestricted in terms of use, does not require a written letter of permission for implementation and clinical use, and asks that the copyright line appear at the bottom of pocket cards and other educational materials. A hospital's own flowsheet, pocket card, or training deck sits inside that permission with the copyright line on it. For anything else the site says one thing: for other uses, contact [email protected] for permission. As of September 2026 Vanderbilt publishes no separate terms for EHR vendors, publications, commercial products, or public web tools, and it has not said in terms whether a health system's own EHR build counts as implementation, so a cautious team confirms with the center and keeps the copyright line in the build, while a vendor product, a paid app, a public calculator, or a published reproduction of the worksheet is an other use to clear with the center first. Free, attributed, or non-commercial does not by itself create a license, and the CAM-ICU is not public domain. Two neighbors have their own owners: the RASS on the pocket card is Virginia Commonwealth University's copyrighted scale with its own free-use terms, and the original CAM belongs to Dr. Inouye's Hospital Elder Life Program. This page and its templates stay on the right side of that line: the feature logic in original words, and no worksheet, stimuli, thresholds, or manual text.
Yes. Give it the facts (the RASS and time, the baseline and its source, each feature assessed and where the assessment stopped, the result and subtype or the reason it could not be done, the contributors reviewed, the actions taken, who was notified and what they decided, and the prior result) and it drafts the full entry: gate, features, result, contributors, action, notification, and the handoff line, ready for your review. It can also check a finished note for a negative at an unassessable RASS, a UTA with no reason, a result with no feature trail, fluctuation lost at a calm morning, a positive with nothing after it, or a nursing note that states the diagnosis. BastionGPT is HIPAA-compliant with a signed BAA on every plan, and your data is never used to train models.
The instrument facts and compliance claims on this page trace to these sources, last verified September 2026:
Educational content, not legal or billing advice. Sample notes are fictional. Follow your organization's policies and your board, payer, and jurisdiction requirements.