Head to head

Comparing the main delirium screening tools

Most clinicians only need a handful of instruments. This page sets the best-studied screening and diagnostic tools side by side, using pooled meta-analytic accuracy where it exists and clearly labelling figures that come from a single original study.

Disclosure

Estimates are drawn from cited systematic reviews and meta-analyses. Professor MacLullich is a co-author of the 4AT meta-analysis (Tieges 2021) and the real-world implementation review (Penfold 2024); this is disclosed because those sources inform prominent site content. No instrument is ranked or pre-selected. See Methodology and Disclosures.

ToolBest-fit settingItemsTime SensitivitySpecificityEvidence for the figure
CAM
Confusion Assessment Method		 General / medical wards4~5 min 82–94%89–99% Pooled across syntheses (Wei 2008; Shi 2013)
4AT
4 ’A’s Test		 General ward, ED, perioperative4~2 min 88%88% Pooled, 17 studies (Tieges 2021)
3D-CAM General / medical wards22~3 min 92%95% Pooled, 7 studies (Ma 2023)
bCAM + DTS Emergency department2 + 4<2 min DTS 98% (rule-out)bCAM 96–97% (rule-in) Original study (Han 2013), 2-step approach
CAM-ICU Intensive care (ventilated)4~2 min 80–81%96–98% Pooled (Gusmão-Flores 2012; Shi 2013)
ICDSC Intensive care (screening)8<5 min 74%82% Pooled, 4 studies (Gusmão-Flores 2012)
Nu-DESC Ward / oncology monitoring5~1 min 86%87% Original study (Gaudreau 2005)
UB-2 General (first-step screen)2<1 min 93%64% Original study (Fick 2015)

Pooled figures are bivariate random-effects estimates from the cited meta-analyses; ranges reflect different syntheses, populations and cut-offs. Single-study figures are labelled as such. Sensitivity and specificity trade off against each other: a highly sensitive rule-out screen (e.g. DTS) is deliberately less specific.

The classic question

4AT vs CAM

Both are strong, well-validated tools built for different jobs. Neither is universally “better”; the right choice depends on the task, the setting, and how much rater training you can guarantee.

CAM
Diagnostic algorithm

Strengths

  • The reference algorithm — the most widely used and studied instrument, and the basis of many others (3D-CAM, bCAM, CAM-ICU).
  • Maps directly to DSM features, so it is well suited to structured diagnosis and research; high pooled specificity (up to 99%).
  • Extensively translated and embedded in guidelines and training programmes worldwide.

Considerations

  • Performs best when scored during a formal cognitive assessment by a trained rater; its developers recommend training (Wei 2008).
  • Pooled sensitivity varies by synthesis and setting (82–94%), and it takes a little longer than ultra-brief screens.
4AT
Ultra-brief screen

Strengths

  • No specific training required and ~2 minutes, with an observational item so it can be completed in drowsy or uncooperative patients.
  • Incorporates arousal and a brief cognitive screen; pooled sensitivity 88%, specificity 88% across 17 studies (Tieges 2021).
  • Recommended by NICE (2023) as an option for most settings.

Considerations

  • It is a screening tool, not a diagnostic algorithm: a positive 4AT still requires a full clinical assessment.
  • Like all brief screens it trades some diagnostic detail for speed, and does not itself specify motor subtype or severity.

How to choose between them

Match the tool to the rater and the task. Where a structured, feature-based diagnosis is the goal and trained assessors are available, a CAM-based instrument (CAM or 3D-CAM) is a natural fit. Where the goal is universal, staff-delivered case-finding at scale — and consistent training cannot be guaranteed — an ultra-brief tool that needs none (such as the 4AT) is often more practical. In many services the two are used together: a brief screen at the front door, and a CAM-based assessment when the screen is positive.

In the ICU

CAM-ICU vs ICDSC

Both are recommended for delirium screening in critically ill adults and appear in major critical-care guidance. In pooled analyses the CAM-ICU shows high specificity (96–98%) with good sensitivity (80–81%), making it a strong rule-in assessment; the ICDSC has moderate sensitivity (74%) and good specificity (82%), and can be scored from routine observation across a shift without a formal patient encounter. Many units use the ICDSC for continuous nurse screening and the CAM-ICU for structured assessment.

Screening is not diagnosis

Every tool here is an aid to detection, not a substitute for clinical judgement. A positive screen should prompt a full assessment for delirium and its causes; a negative screen in a patient you are worried about should not reassure you. Accuracy also varies with setting, cut-off, and the training of the person administering the tool.

References

Accuracy figures and citations were verified against PubMed (July 2026). Data retrieved from PubMed; DOIs link to the original articles.

  1. Wei LA, Fearing MA, Sternberg EJ, Inouye SK. The Confusion Assessment Method: a systematic review of current usage. J Am Geriatr Soc 2008;56(5):823-30. PubMed · DOI
  2. Shi Q, Warren L, Saposnik G, MacDermid JC. Confusion assessment method: a systematic review and meta-analysis of diagnostic accuracy. Neuropsychiatr Dis Treat 2013;9:1359-70. PubMed · DOI
  3. Inouye SK, van Dyck CH, Alessi CA, et al. Clarifying confusion: the confusion assessment method. Ann Intern Med 1990;113(12):941-8. PubMed · DOI
  4. Bellelli G, Morandi A, Davis DHJ, et al. Validation of the 4AT. Age Ageing 2014;43(4):496-502. PubMed · DOI
  5. Tieges Z, MacLullich AMJ, Anand A, et al. Diagnostic accuracy of the 4AT: systematic review and meta-analysis. Age Ageing 2021;50(3):733-743. PubMed · DOI
  6. Marcantonio ER, Ngo LH, O’Connor M, et al. 3D-CAM. Ann Intern Med 2014;161(8):554-61. PubMed · DOI
  7. Ma R, Zhao J, Li C, et al. Diagnostic accuracy of the 3D-CAM: systematic review and meta-analysis. Age Ageing 2023;52(5):afad074. PubMed · DOI
  8. Han JH, Wilson A, Vasilevskis EE, et al. Delirium triage screen and brief CAM. Ann Emerg Med 2013;62(5):457-465. PubMed · DOI
  9. Gusmão-Flores D, Salluh JIF, Chalhub RÁ, Quarantini LC. CAM-ICU and ICDSC: systematic review and meta-analysis. Crit Care 2012;16(4):R115. PubMed · DOI
  10. Gaudreau JD, Gagnon P, Harel F, et al. Nu-DESC. J Pain Symptom Manage 2005;29(4):368-75. PubMed · DOI
  11. Fick DM, Inouye SK, Guess J, et al. Ultra-brief 2-item bedside test. J Hosp Med 2015;10(10):645-50. PubMed · DOI